prometheus: Add detailed metrics

Signed-off-by: Julien Pivotto <roidelapluie@inuits.eu>

.github/ISSUE_TEMPLATE.md

@@ -0,0 +1,9 @@
+## Versions:
+
+* mgmt version (eg: `mgmt --version`):
+
+* operating system/distribution (eg: `uname -a`):
+
+* golang version (eg: `go version`):
+
+## Description:

.gitignore

@@ -2,7 +2,6 @@
 .omv/
 .ssh/
 .vagrant/
-mgmt-documentation.pdf
 old/
 tmp/
 *_stringer.go

.gitmodules

@@ -13,3 +13,6 @@
 [submodule "vendor/github.com/purpleidea/go-systemd"]
 	path = vendor/github.com/purpleidea/go-systemd
 	url = https://github.com/purpleidea/go-systemd
+[submodule "vendor/honnef.co/go/augeas"]
+	path = vendor/honnef.co/go/augeas
+	url = https://github.com/dominikh/go-augeas/

.travis.yml

@@ -1,18 +1,22 @@
 language: go
 go:
-  - 1.6
-  - 1.7
+  - 1.6.x
+  - 1.7.x
+  - 1.8.x
   - tip
+go_import_path: github.com/purpleidea/mgmt
 sudo: true
 dist: trusty
-before_install: 'git fetch --unshallow'
+before_install:
+  - sudo apt update
+  - git fetch --unshallow
 install: 'make deps'
 script: 'make test'
 matrix:
   fast_finish: true
   allow_failures:
     - go: tip
-    - go: 1.7
+    - go: 1.8.x
 notifications:
   irc:
     channels:
@@ -25,4 +29,7 @@ notifications:
     use_notice: false
     skip_join: false
   email:
-    - travis-ci@shubin.ca
+    recipients:
+      - travis-ci@shubin.ca
+    on_failure: change
+    on_success: change

AUTHORS

@@ -4,5 +4,7 @@ If you appreciate the work of one of the contributors, thank them a beverage!
 For a more exhaustive list please run: git log --format='%aN' | sort -u
 This list is sorted alphabetically by first name.
 
+Felix Frank
 James Shubin
+Julien Pivotto
 Paul Morgan

COPYRIGHT

@@ -1,5 +1,5 @@
 Mgmt
-Copyright (C) 2013-2016+ James Shubin and the project contributors
+Copyright (C) 2013-2017+ James Shubin and the project contributors
 Written by James Shubin <james@shubin.ca> and the project contributors
 
 This program is free software: you can redistribute it and/or modify

Makefile

@@ -1,5 +1,5 @@
 # Mgmt
-# Copyright (C) 2013-2016+ James Shubin and the project contributors
+# Copyright (C) 2013-2017+ James Shubin and the project contributors
 # Written by James Shubin <james@shubin.ca> and the project contributors
 #
 # This program is free software: you can redistribute it and/or modify
@@ -37,6 +37,11 @@ RPM = rpmbuild/RPMS/$(PROGRAM)-$(VERSION)-$(RELEASE).$(ARCH).rpm
 USERNAME := $(shell cat ~/.config/copr 2>/dev/null | grep username | awk -F '=' '{print $$2}' | tr -d ' ')
 SERVER = 'dl.fedoraproject.org'
 REMOTE_PATH = 'pub/alt/$(USERNAME)/$(PROGRAM)'
+ifneq ($(GOTAGS),)
+    BUILD_FLAGS = -tags '$(GOTAGS)'
+endif
+
+default: build
 
 #
 #	art
@@ -105,9 +110,9 @@ $(PROGRAM): main.go
 	@echo "Building: $(PROGRAM), version: $(SVERSION)..."
 ifneq ($(OLDGOLANG),)
 	@# avoid equals sign in old golang versions eg in: -X foo=bar
-	time go build -ldflags "-X main.program $(PROGRAM) -X main.version $(SVERSION)" -o $(PROGRAM);
+	time go build -ldflags "-X main.program $(PROGRAM) -X main.version $(SVERSION)" -o $(PROGRAM) $(BUILD_FLAGS);
 else
-	time go build -ldflags "-X main.program=$(PROGRAM) -X main.version=$(SVERSION)" -o $(PROGRAM);
+	time go build -i -ldflags "-X main.program=$(PROGRAM) -X main.version=$(SVERSION)" -o $(PROGRAM) $(BUILD_FLAGS);
 endif
 
 $(PROGRAM).static: main.go
@@ -115,9 +120,9 @@ $(PROGRAM).static: main.go
 	go generate
 ifneq ($(OLDGOLANG),)
 	@# avoid equals sign in old golang versions eg in: -X foo=bar
-	go build -a -installsuffix cgo -tags netgo -ldflags '-extldflags "-static" -X main.program $(PROGRAM) -X main.version $(SVERSION)' -o $(PROGRAM).static;
+	go build -a -installsuffix cgo -tags netgo -ldflags '-extldflags "-static" -X main.program $(PROGRAM) -X main.version $(SVERSION)' -o $(PROGRAM).static $(BUILD_FLAGS);
 else
-	go build -a -installsuffix cgo -tags netgo -ldflags '-extldflags "-static" -X main.program=$(PROGRAM) -X main.version=$(SVERSION)' -o $(PROGRAM).static;
+	go build -a -installsuffix cgo -tags netgo -ldflags '-extldflags "-static" -X main.program=$(PROGRAM) -X main.version=$(SVERSION)' -o $(PROGRAM).static $(BUILD_FLAGS);
 endif
 
 clean:
@@ -138,8 +143,8 @@ format: gofmt yamlfmt
 
 docs: $(PROGRAM)-documentation.pdf
 
-$(PROGRAM)-documentation.pdf: DOCUMENTATION.md
-	pandoc DOCUMENTATION.md -o '$(PROGRAM)-documentation.pdf'
+$(PROGRAM)-documentation.pdf: docs/documentation.md
+	pandoc docs/documentation.md -o docs/'$(PROGRAM)-documentation.pdf'
 
 #
 #	build aliases
@@ -183,7 +188,7 @@ $(SRPM): $(SPEC) $(SOURCE)
 #
 $(SPEC): rpmbuild/ spec.in
 	@echo Running templater...
-	#cat spec.in > $(SPEC)
+	cat spec.in > $(SPEC)
 	sed -e s/__PROGRAM__/$(PROGRAM)/g -e s/__VERSION__/$(VERSION)/g -e s/__RELEASE__/$(RELEASE)/g < spec.in > $(SPEC)
 	# append a changelog to the .spec file
 	git log --format="* %cd %aN <%aE>%n- (%h) %s%d%n" --date=local | sed -r 's/[0-9]+:[0-9]+:[0-9]+ //' >> $(SPEC)

README.md

@@ -2,18 +2,20 @@
 
 [![mgmt!](art/mgmt.png)](art/)
 
-[![Go Report Card](https://goreportcard.com/badge/github.com/purpleidea/mgmt)](https://goreportcard.com/report/github.com/purpleidea/mgmt)
-[![Build Status](https://secure.travis-ci.org/purpleidea/mgmt.png?branch=master)](http://travis-ci.org/purpleidea/mgmt)
-[![Documentation](https://img.shields.io/docs/markdown.png)](DOCUMENTATION.md)
-[![GoDoc](https://godoc.org/github.com/purpleidea/mgmt?status.svg)](https://godoc.org/github.com/purpleidea/mgmt)
-[![IRC](https://img.shields.io/irc/%23mgmtconfig.png)](https://webchat.freenode.net/?channels=#mgmtconfig)
-[![Jenkins](https://img.shields.io/jenkins/status.png)](https://ci.centos.org/job/purpleidea-mgmt/)
-[![COPR](https://img.shields.io/copr/builds.png)](https://copr.fedoraproject.org/coprs/purpleidea/mgmt/)
-[![arch](https://img.shields.io/arch/aur.png)](https://aur.archlinux.org/packages/mgmt/)
+[![Go Report Card](https://goreportcard.com/badge/github.com/purpleidea/mgmt?style=flat-square)](https://goreportcard.com/report/github.com/purpleidea/mgmt)
+[![Build Status](https://img.shields.io/travis/purpleidea/mgmt/master.svg?style=flat-square)](http://travis-ci.org/purpleidea/mgmt)
+[![GoDoc](https://img.shields.io/badge/godoc-reference-5272B4.svg?style=flat-square)](https://godoc.org/github.com/purpleidea/mgmt)
+[![IRC](https://img.shields.io/badge/irc-%23mgmtconfig-brightgreen.svg?style=flat-square)](https://webchat.freenode.net/?channels=#mgmtconfig)
+[![Jenkins](https://img.shields.io/badge/jenkins-status-brightgreen.svg?style=flat-square)](https://ci.centos.org/job/purpleidea-mgmt/)
 
 ## Community:
-Come join us on IRC in [#mgmtconfig](https://webchat.freenode.net/?channels=#mgmtconfig) on Freenode!
-You may like the [#mgmtconfig](https://twitter.com/hashtag/mgmtconfig) hashtag if you're on [Twitter](https://twitter.com/#!/purpleidea).
+Come join us in the `mgmt` community!
+
+| Medium | Link |
+|---|---|---|
+| IRC | [#mgmtconfig](https://webchat.freenode.net/?channels=#mgmtconfig) on Freenode |
+| Twitter | [@mgmtconfig](https://twitter.com/mgmtconfig) & [#mgmtconfig](https://twitter.com/hashtag/mgmtconfig) |
+| Mailing list | [mgmtconfig-list@redhat.com](https://www.redhat.com/mailman/listinfo/mgmtconfig-list) |
 
 ## Status:
 Mgmt is a fairly new project.
@@ -21,35 +23,21 @@ We're working towards being minimally useful for production environments.
 We aren't feature complete for what we'd consider a 1.x release yet.
 With your help you'll be able to influence our design and get us there sooner!
 
-## Questions:
-Please join the [#mgmtconfig](https://webchat.freenode.net/?channels=#mgmtconfig) IRC community!
-If you have a well phrased question that might benefit others, consider asking it by sending a patch to the documentation [FAQ](https://github.com/purpleidea/mgmt/blob/master/DOCUMENTATION.md#usage-and-frequently-asked-questions) section. I'll merge your question, and a patch with the answer!
-
-## Quick start:
-* Make sure you have golang version 1.6 or greater installed.
-* If you do not have a GOPATH yet, create one and export it:
-```
-mkdir $HOME/gopath
-export GOPATH=$HOME/gopath
-```
-* You might also want to add the GOPATH to your `~/.bashrc` or `~/.profile`.
-* For more information you can read the [GOPATH documentation](https://golang.org/cmd/go/#hdr-GOPATH_environment_variable).
-* Next download the mgmt code base, and switch to that directory:
-```
-go get -u github.com/purpleidea/mgmt
-cd $GOPATH/src/github.com/purpleidea/mgmt
-```
-* Get the remaining golang deps with `go get ./...`, or run `make deps` if you're comfortable with how we install them.
-* Run `make build` to get a freshly built `mgmt` binary.
-* Run `time ./mgmt run --yaml examples/graph0.yaml --converged-timeout=5 --tmp-prefix` to try out a very simple example!
-* To run continuously in the default mode of operation, omit the `--converged-timeout` option.
-* Have fun hacking on our future technology!
-
-## Examples:
-Please look in the [examples/](examples/) folder for more examples!
-
 ## Documentation:
-Please see: the manually created [DOCUMENTATION.md](DOCUMENTATION.md) (also available as [PDF](https://pdfdoc-purpleidea.rhcloud.com/pdf/https://github.com/purpleidea/mgmt/blob/master/DOCUMENTATION.md)) and the automatically generated [GoDoc documentation](https://godoc.org/github.com/purpleidea/mgmt).
+Please read, enjoy and help improve our documentation!
+
+| Documentation | Additional Notes |
+|---|---|
+| [general documentation](docs/documentation.md) | for everyone |
+| [quick start guide](docs/quick-start-guide.md) | for mgmt developers |
+| [resource guide](docs/resource-guide.md) | for mgmt developers |
+| [godoc API reference](https://godoc.org/github.com/purpleidea/mgmt) | for mgmt developers |
+| [prometheus guide](docs/prometheus.md) | for everyone |
+| [puppet guide](docs/puppet-guide.md) | for puppet sysadmins |
+
+## Questions:
+Please ask in the [community](#community)!
+If you have a well phrased question that might benefit others, consider asking it by sending a patch to the documentation [FAQ](https://github.com/purpleidea/mgmt/blob/master/docs/documentation.md#usage-and-frequently-asked-questions) section. I'll merge your question, and a patch with the answer!
 
 ## Roadmap:
 Please see: [TODO.md](TODO.md) for a list of upcoming work and TODO items.
@@ -61,52 +49,35 @@ Please set the `DEBUG` constant in [main.go](https://github.com/purpleidea/mgmt/
 Bonus points if you provide a [shell](https://github.com/purpleidea/mgmt/tree/master/test/shell) or [OMV](https://github.com/purpleidea/mgmt/tree/master/test/omv) reproducible test case.
 Feel free to read my article on [debugging golang programs](https://ttboj.wordpress.com/2016/02/15/debugging-golang-programs/).
 
-## Dependencies:
-* golang 1.6 or higher (required, available in most distros)
-* golang libraries (required, available with `go get`)
-```
-go get github.com/coreos/etcd/client
-go get gopkg.in/yaml.v2
-go get gopkg.in/fsnotify.v1
-go get github.com/urfave/cli
-go get github.com/coreos/go-systemd/dbus
-go get github.com/coreos/go-systemd/util
-go get github.com/coreos/pkg/capnslog
-go get github.com/rgbkrk/libvirt-go
-```
-* stringer (optional for building), available as a package on some platforms, otherwise via `go get`
-```
-go get golang.org/x/tools/cmd/stringer
-```
-* pandoc (optional, for building a pdf of the documentation)
-* graphviz (optional, for building a visual representation of the graph)
-
 ## Patches:
 We'd love to have your patches! Please send them by email, or as a pull request.
 
 ## On the web:
-* James Shubin; blog: [Next generation configuration mgmt](https://ttboj.wordpress.com/2016/01/18/next-generation-configuration-mgmt/)
-* James Shubin; video: [Introductory recording from DevConf.cz 2016](https://www.youtube.com/watch?v=GVhpPF0j-iE&html5=1)
-* James Shubin; video: [Introductory recording from CfgMgmtCamp.eu 2016](https://www.youtube.com/watch?v=fNeooSiIRnA&html5=1)
-* Julian Dunn; video: [On mgmt at CfgMgmtCamp.eu 2016](https://www.youtube.com/watch?v=kfF9IATUask&t=1949&html5=1)
-* Walter Heck; slides: [On mgmt at CfgMgmtCamp.eu 2016](http://www.slideshare.net/olindata/configuration-management-time-for-a-4th-generation/3)
-* Marco Marongiu; blog: [On mgmt](http://syslog.me/2016/02/15/leap-or-die/)
-* Felix Frank; blog: [From Catalog To Mgmt (on puppet to mgmt "transpiling")](https://ffrank.github.io/features/2016/02/18/from-catalog-to-mgmt/)
-* James Shubin; blog: [Automatic edges in mgmt (...and the pkg resource)](https://ttboj.wordpress.com/2016/03/14/automatic-edges-in-mgmt/)
-* James Shubin; blog: [Automatic grouping in mgmt](https://ttboj.wordpress.com/2016/03/30/automatic-grouping-in-mgmt/)
-* John Arundel; tweet: [“Puppet’s days are numbered.”](https://twitter.com/bitfield/status/732157519142002688)
-* Felix Frank; blog: [Puppet, Meet Mgmt (on puppet to mgmt internals)](https://ffrank.github.io/features/2016/06/12/puppet,-meet-mgmt/)
-* Felix Frank; blog: [Puppet Powered Mgmt (puppet to mgmt tl;dr)](https://ffrank.github.io/features/2016/06/19/puppet-powered-mgmt/)
-* James Shubin; blog: [Automatic clustering in mgmt](https://ttboj.wordpress.com/2016/06/20/automatic-clustering-in-mgmt/)
-* James Shubin; video: [Recording from CoreOSFest 2016](https://www.youtube.com/watch?v=KVmDCUA42wc&html5=1)
-* James Shubin; video: [Recording from DebConf16](http://meetings-archive.debian.net/pub/debian-meetings/2016/debconf16/Next_Generation_Config_Mgmt.webm) ([Slides](https://annex.debconf.org//debconf-share/debconf16/slides/15-next-generation-config-mgmt.pdf))
-* Felix Frank; blog: [Edging It All In (puppet and mgmt edges)](https://ffrank.github.io/features/2016/07/12/edging-it-all-in/)
-* Felix Frank; blog: [Translating All The Things (puppet to mgmt translation warnings)](https://ffrank.github.io/features/2016/08/19/translating-all-the-things/)
-* James Shubin; video: [Recording from systemd.conf 2016](https://www.youtube.com/watch?v=jB992Zb3nH0&html5=1)
-* James Shubin; blog: [Remote execution in mgmt](https://ttboj.wordpress.com/2016/10/07/remote-execution-in-mgmt/)
-* James Shubin; video: [Recording from High Load Strategy 2016](https://vimeo.com/191493409)
-* James Shubin; video: [Recording from NLUUG 2016](https://www.youtube.com/watch?v=MmpwOQAb_SE&html5=1)
-* James Shubin; blog: [Send/Recv in mgmt](https://ttboj.wordpress.com/2016/12/07/sendrecv-in-mgmt/)
+| Author | Format | Subject |
+|---|---|---|
+| James Shubin | blog | [Next generation configuration mgmt](https://ttboj.wordpress.com/2016/01/18/next-generation-configuration-mgmt/) |
+| James Shubin | video | [Introductory recording from DevConf.cz 2016](https://www.youtube.com/watch?v=GVhpPF0j-iE&html5=1) |
+| James Shubin | video | [Introductory recording from CfgMgmtCamp.eu 2016](https://www.youtube.com/watch?v=fNeooSiIRnA&html5=1) |
+| Julian Dunn | video | [On mgmt at CfgMgmtCamp.eu 2016](https://www.youtube.com/watch?v=kfF9IATUask&t=1949&html5=1) |
+| Walter Heck | slides | [On mgmt at CfgMgmtCamp.eu 2016](http://www.slideshare.net/olindata/configuration-management-time-for-a-4th-generation/3) |
+| Marco Marongiu | blog | [On mgmt](http://syslog.me/2016/02/15/leap-or-die/) |
+| Felix Frank | blog | [From Catalog To Mgmt (on puppet to mgmt "transpiling")](https://ffrank.github.io/features/2016/02/18/from-catalog-to-mgmt/) |
+| James Shubin | blog | [Automatic edges in mgmt (...and the pkg resource)](https://ttboj.wordpress.com/2016/03/14/automatic-edges-in-mgmt/) |
+| James Shubin | blog | [Automatic grouping in mgmt](https://ttboj.wordpress.com/2016/03/30/automatic-grouping-in-mgmt/) |
+| John Arundel | tweet | [“Puppet’s days are numbered.”](https://twitter.com/bitfield/status/732157519142002688) |
+| Felix Frank | blog | [Puppet, Meet Mgmt (on puppet to mgmt internals)](https://ffrank.github.io/features/2016/06/12/puppet,-meet-mgmt/) |
+| Felix Frank | blog | [Puppet Powered Mgmt (puppet to mgmt tl;dr)](https://ffrank.github.io/features/2016/06/19/puppet-powered-mgmt/) |
+| James Shubin | blog | [Automatic clustering in mgmt](https://ttboj.wordpress.com/2016/06/20/automatic-clustering-in-mgmt/) |
+| James Shubin | video | [Recording from CoreOSFest 2016](https://www.youtube.com/watch?v=KVmDCUA42wc&html5=1) |
+| James Shubin | video | [Recording from DebConf16](http://meetings-archive.debian.net/pub/debian-meetings/2016/debconf16/Next_Generation_Config_Mgmt.webm) ([Slides](https://annex.debconf.org//debconf-share/debconf16/slides/15-next-generation-config-mgmt.pdf)) |
+| Felix Frank | blog | [Edging It All In (puppet and mgmt edges)](https://ffrank.github.io/features/2016/07/12/edging-it-all-in/) |
+| Felix Frank | blog | [Translating All The Things (puppet to mgmt translation warnings)](https://ffrank.github.io/features/2016/08/19/translating-all-the-things/) |
+| James Shubin | video | [Recording from systemd.conf 2016](https://www.youtube.com/watch?v=jB992Zb3nH0&html5=1) |
+| James Shubin | blog | [Remote execution in mgmt](https://ttboj.wordpress.com/2016/10/07/remote-execution-in-mgmt/) |
+| James Shubin | video | [Recording from High Load Strategy 2016](https://vimeo.com/191493409) |
+| James Shubin | video | [Recording from NLUUG 2016](https://www.youtube.com/watch?v=MmpwOQAb_SE&html5=1) |
+| James Shubin | blog | [Send/Recv in mgmt](https://ttboj.wordpress.com/2016/12/07/sendrecv-in-mgmt/) |
+| James Shubin | blog | [Metaparameters in mgmt](https://ttboj.wordpress.com/2017/03/01/metaparameters-in-mgmt/) |
 
 ##
 

TODO.md

@@ -1,16 +1,16 @@
 # TODO
 If you're looking for something to do, look here!
 Let us know if you're working on one of the items.
+If you'd like something to work on, ping @purpleidea and I'll create an issue
+tailored especially for you! Just let me know your approximate golang skill
+level and how many hours you'd like to spend on the patch.
 
 ## Package resource
 - [ ] getfiles support on debian [bug](https://github.com/hughsie/PackageKit/issues/118)
 - [ ] directory info on fedora [bug](https://github.com/hughsie/PackageKit/issues/117)
 - [ ] dnf blocker [bug](https://github.com/hughsie/PackageKit/issues/110)
-- [ ] install signal blocker [bug](https://github.com/hughsie/PackageKit/issues/109)
 
-## File resource [bug](https://github.com/purpleidea/mgmt/issues/13) [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
-- [ ] chown/chmod support [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
-- [ ] user/group support [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
+## File resource [bug](https://github.com/purpleidea/mgmt/issues/64) [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 - [ ] recurse limit support [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 - [ ] fanotify support [bug](https://github.com/go-fsnotify/fsnotify/issues/114)
 
@@ -21,7 +21,6 @@ Let us know if you're working on one of the items.
 - [ ] base resource improvements
 
 ## Timer resource
-- [ ] reset on recompile
 - [ ] increment algorithm (linear, exponential, etc...) [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 
 ## User/Group resource
@@ -29,7 +28,7 @@ Let us know if you're working on one of the items.
 - [ ] automatic edges to file resource [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 
 ## Virt (libvirt) resource
-- [ ] base resource [bug](https://github.com/purpleidea/mgmt/issues/25)
+- [ ] base resource improvements [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 
 ## Net (systemd-networkd) resource
 - [ ] base resource
@@ -44,7 +43,7 @@ Let us know if you're working on one of the items.
 - [ ] base resource [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 
 ## Http resource
-- [ ] base resource
+- [ ] base resource [:heart:](https://github.com/purpleidea/mgmt/labels/mgmtlove)
 
 ## Etcd improvements
 - [ ] fix embedded etcd master race
@@ -52,6 +51,9 @@ Let us know if you're working on one of the items.
 ## Torrent/dht file transfer
 - [ ] base plumbing
 
+## GPG/Auth improvements
+- [ ] base plumbing
+
 ## Language improvements
 - [ ] language design
 - [ ] lexer/parser

Vagrantfile

@@ -0,0 +1,38 @@
+Vagrant.configure(2) do |config|
+	config.ssh.forward_agent = true
+	config.ssh.username = 'vagrant'
+	config.vm.network "private_network", ip: "192.168.219.2"
+
+	config.vm.synced_folder ".", "/vagrant", disabled: true
+
+	config.vm.define "mgmt-dev" do |instance|
+		instance.vm.box = "fedora/24-cloud-base"
+	end
+
+	config.vm.provider "virtualbox" do |v|
+		v.memory = 1536
+		v.cpus = 2
+	end
+
+	config.vm.provision "file", source: "vagrant/motd", destination: ".motd"
+	config.vm.provision "shell", inline: "cp ~vagrant/.motd /etc/motd"
+
+	config.vm.provision "file", source: "vagrant/mgmt.bashrc", destination: ".mgmt.bashrc"
+	config.vm.provision "file", source: "~/.gitconfig", destination: ".gitconfig"
+
+	# copied from make-deps.sh (with added git)
+	config.vm.provision "shell", inline: "dnf install -y libvirt-devel golang golang-googlecode-tools-stringer hg git"
+
+	# set up vagrant home
+	script = <<-SCRIPT
+		grep -q 'mgmt\.bashrc' ~/.bashrc || echo '. ~/.mgmt.bashrc' >>~/.bashrc
+		. ~/.mgmt.bashrc
+		go get -u github.com/purpleidea/mgmt
+		cd ~/gopath/src/github.com/purpleidea/mgmt
+		make deps
+	SCRIPT
+	config.vm.provision "shell" do |shell|
+		shell.privileged = false
+		shell.inline = script
+	end
+end

converger/converger.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -29,24 +29,24 @@ import (
 // TODO: we could make a new function that masks out the state of certain
 // UID's, but at the moment the new Timer code has obsoleted the need...
 
-// Converger is the general interface for implementing a convergence watcher
+// Converger is the general interface for implementing a convergence watcher.
 type Converger interface { // TODO: need a better name
-	Register() ConvergerUID
-	IsConverged(ConvergerUID) bool         // is the UID converged ?
-	SetConverged(ConvergerUID, bool) error // set the converged state of the UID
-	Unregister(ConvergerUID)
+	Register() UID
+	IsConverged(UID) bool         // is the UID converged ?
+	SetConverged(UID, bool) error // set the converged state of the UID
+	Unregister(UID)
 	Start()
 	Pause()
 	Loop(bool)
-	ConvergedTimer(ConvergerUID) <-chan time.Time
+	ConvergedTimer(UID) <-chan time.Time
 	Status() map[uint64]bool
 	Timeout() int                // returns the timeout that this was created with
 	SetStateFn(func(bool) error) // sets the stateFn
 }
 
-// ConvergerUID is the interface resources can use to notify with if converged
-// you'll need to use part of the Converger interface to Register initially too
-type ConvergerUID interface {
+// UID is the interface resources can use to notify with if converged. You'll
+// need to use part of the Converger interface to Register initially too.
+type UID interface {
 	ID() uint64   // get Id
 	Name() string // get a friendly name
 	SetName(string)
@@ -61,7 +61,7 @@ type ConvergerUID interface {
 	StopTimer() error
 }
 
-// converger is an implementation of the Converger interface
+// converger is an implementation of the Converger interface.
 type converger struct {
 	timeout   int              // must be zero (instant) or greater seconds to run
 	stateFn   func(bool) error // run on converged state changes with state bool
@@ -73,17 +73,18 @@ type converger struct {
 	status    map[uint64]bool
 }
 
-// convergerUID is an implementation of the ConvergerUID interface
-type convergerUID struct {
+// cuid is an implementation of the UID interface.
+type cuid struct {
 	converger Converger
 	id        uint64
 	name      string // user defined, friendly name
 	mutex     sync.Mutex
 	timer     chan struct{}
 	running   bool // is the above timer running?
+	wg        sync.WaitGroup
 }
 
-// NewConverger builds a new converger struct
+// NewConverger builds a new converger struct.
 func NewConverger(timeout int, stateFn func(bool) error) *converger {
 	return &converger{
 		timeout: timeout,
@@ -95,13 +96,13 @@ func NewConverger(timeout int, stateFn func(bool) error) *converger {
 	}
 }
 
-// Register assigns a ConvergerUID to the caller
-func (obj *converger) Register() ConvergerUID {
+// Register assigns a UID to the caller.
+func (obj *converger) Register() UID {
 	obj.mutex.Lock()
 	defer obj.mutex.Unlock()
 	obj.lastid++
 	obj.status[obj.lastid] = false // initialize as not converged
-	return &convergerUID{
+	return &cuid{
 		converger: obj,
 		id:        obj.lastid,
 		name:      fmt.Sprintf("%d", obj.lastid), // some default
@@ -110,28 +111,28 @@ func (obj *converger) Register() ConvergerUID {
 	}
 }
 
-// IsConverged gets the converged status of a uid
-func (obj *converger) IsConverged(uid ConvergerUID) bool {
+// IsConverged gets the converged status of a uid.
+func (obj *converger) IsConverged(uid UID) bool {
 	if !uid.IsValid() {
-		panic(fmt.Sprintf("Id of ConvergerUID(%s) is nil!", uid.Name()))
+		panic(fmt.Sprintf("the ID of UID(%s) is nil", uid.Name()))
 	}
 	obj.mutex.RLock()
 	isConverged, found := obj.status[uid.ID()] // lookup
 	obj.mutex.RUnlock()
 	if !found {
-		panic("Id of ConvergerUID is unregistered!")
+		panic("the ID of UID is unregistered")
 	}
 	return isConverged
 }
 
-// SetConverged updates the converger with the converged state of the UID
-func (obj *converger) SetConverged(uid ConvergerUID, isConverged bool) error {
+// SetConverged updates the converger with the converged state of the UID.
+func (obj *converger) SetConverged(uid UID, isConverged bool) error {
 	if !uid.IsValid() {
-		return fmt.Errorf("Id of ConvergerUID(%s) is nil!", uid.Name())
+		return fmt.Errorf("the ID of UID(%s) is nil", uid.Name())
 	}
 	obj.mutex.Lock()
 	if _, found := obj.status[uid.ID()]; !found {
-		panic("Id of ConvergerUID is unregistered!")
+		panic("the ID of UID is unregistered")
 	}
 	obj.status[uid.ID()] = isConverged // set
 	obj.mutex.Unlock()                 // unlock *before* poke or deadlock!
@@ -143,7 +144,7 @@ func (obj *converger) SetConverged(uid ConvergerUID, isConverged bool) error {
 	return nil
 }
 
-// isConverged returns true if *every* registered uid has converged
+// isConverged returns true if *every* registered uid has converged.
 func (obj *converger) isConverged() bool {
 	obj.mutex.RLock() // take a read lock
 	defer obj.mutex.RUnlock()
@@ -155,10 +156,10 @@ func (obj *converger) isConverged() bool {
 	return true
 }
 
-// Unregister dissociates the ConvergedUID from the converged checking
-func (obj *converger) Unregister(uid ConvergerUID) {
+// Unregister dissociates the ConvergedUID from the converged checking.
+func (obj *converger) Unregister(uid UID) {
 	if !uid.IsValid() {
-		panic(fmt.Sprintf("Id of ConvergerUID(%s) is nil!", uid.Name()))
+		panic(fmt.Sprintf("the ID of UID(%s) is nil", uid.Name()))
 	}
 	obj.mutex.Lock()
 	uid.StopTimer() // ignore any errors
@@ -167,30 +168,30 @@ func (obj *converger) Unregister(uid ConvergerUID) {
 	uid.InvalidateID()
 }
 
-// Start causes a Converger object to start or resume running
+// Start causes a Converger object to start or resume running.
 func (obj *converger) Start() {
 	obj.control <- true
 }
 
-// Pause causes a Converger object to stop running temporarily
+// Pause causes a Converger object to stop running temporarily.
 func (obj *converger) Pause() { // FIXME: add a sync ACK on pause before return
 	obj.control <- false
 }
 
-// Loop is the main loop for a Converger object; it usually runs in a goroutine
-// TODO: we could eventually have each resource tell us as soon as it converges
-// and then keep track of the time delays here, to avoid callers needing select
+// Loop is the main loop for a Converger object. It usually runs in a goroutine.
+// TODO: we could eventually have each resource tell us as soon as it converges,
+// and then keep track of the time delays here, to avoid callers needing select.
 // NOTE: when we have very short timeouts, if we start before all the resources
-// have joined the map, then it might appears as if we converged before we did!
+// have joined the map, then it might appear as if we converged before we did!
 func (obj *converger) Loop(startPaused bool) {
 	if obj.control == nil {
-		panic("Converger not initialized correctly")
+		panic("converger not initialized correctly")
 	}
 	if startPaused { // start paused without racing
 		select {
 		case e := <-obj.control:
 			if !e {
-				panic("Converger expected true!")
+				panic("converger expected true")
 			}
 		}
 	}
@@ -198,13 +199,13 @@ func (obj *converger) Loop(startPaused bool) {
 		select {
 		case e := <-obj.control: // expecting "false" which means pause!
 			if e {
-				panic("Converger expected false!")
+				panic("converger expected false")
 			}
 			// now i'm paused...
 			select {
 			case e := <-obj.control:
 				if !e {
-					panic("Converger expected true!")
+					panic("converger expected true")
 				}
 				// restart
 				// kick once to refresh the check...
@@ -243,9 +244,9 @@ func (obj *converger) Loop(startPaused bool) {
 	}
 }
 
-// ConvergedTimer adds a timeout to a select call and blocks until then
+// ConvergedTimer adds a timeout to a select call and blocks until then.
 // TODO: this means we could eventually have per resource converged timeouts
-func (obj *converger) ConvergedTimer(uid ConvergerUID) <-chan time.Time {
+func (obj *converger) ConvergedTimer(uid UID) <-chan time.Time {
 	// be clever: if i'm already converged, this timeout should block which
 	// avoids unnecessary new signals being sent! this avoids fast loops if
 	// we have a low timeout, or in particular a timeout == 0
@@ -279,63 +280,65 @@ func (obj *converger) SetStateFn(stateFn func(bool) error) {
 	obj.stateFn = stateFn
 }
 
-// Id returns the unique id of this UID object
-func (obj *convergerUID) ID() uint64 {
+// ID returns the unique id of this UID object.
+func (obj *cuid) ID() uint64 {
 	return obj.id
 }
 
-// Name returns a user defined name for the specific convergerUID.
-func (obj *convergerUID) Name() string {
+// Name returns a user defined name for the specific cuid.
+func (obj *cuid) Name() string {
 	return obj.name
 }
 
-// SetName sets a user defined name for the specific convergerUID.
-func (obj *convergerUID) SetName(name string) {
+// SetName sets a user defined name for the specific cuid.
+func (obj *cuid) SetName(name string) {
 	obj.name = name
 }
 
-// IsValid tells us if the id is valid or has already been destroyed
-func (obj *convergerUID) IsValid() bool {
+// IsValid tells us if the id is valid or has already been destroyed.
+func (obj *cuid) IsValid() bool {
 	return obj.id != 0 // an id of 0 is invalid
 }
 
-// InvalidateID marks the id as no longer valid
-func (obj *convergerUID) InvalidateID() {
+// InvalidateID marks the id as no longer valid.
+func (obj *cuid) InvalidateID() {
 	obj.id = 0 // an id of 0 is invalid
 }
 
-// IsConverged is a helper function to the regular IsConverged method
-func (obj *convergerUID) IsConverged() bool {
+// IsConverged is a helper function to the regular IsConverged method.
+func (obj *cuid) IsConverged() bool {
 	return obj.converger.IsConverged(obj)
 }
 
-// SetConverged is a helper function to the regular SetConverged notification
-func (obj *convergerUID) SetConverged(isConverged bool) error {
+// SetConverged is a helper function to the regular SetConverged notification.
+func (obj *cuid) SetConverged(isConverged bool) error {
 	return obj.converger.SetConverged(obj, isConverged)
 }
 
-// Unregister is a helper function to unregister myself
-func (obj *convergerUID) Unregister() {
+// Unregister is a helper function to unregister myself.
+func (obj *cuid) Unregister() {
 	obj.converger.Unregister(obj)
 }
 
-// ConvergedTimer is a helper around the regular ConvergedTimer method
-func (obj *convergerUID) ConvergedTimer() <-chan time.Time {
+// ConvergedTimer is a helper around the regular ConvergedTimer method.
+func (obj *cuid) ConvergedTimer() <-chan time.Time {
 	return obj.converger.ConvergedTimer(obj)
 }
 
 // StartTimer runs an invisible timer that automatically converges on timeout.
-func (obj *convergerUID) StartTimer() (func() error, error) {
+func (obj *cuid) StartTimer() (func() error, error) {
 	obj.mutex.Lock()
 	if !obj.running {
 		obj.timer = make(chan struct{})
 		obj.running = true
 	} else {
 		obj.mutex.Unlock()
-		return obj.StopTimer, fmt.Errorf("Timer already started!")
+		return obj.StopTimer, fmt.Errorf("timer already started")
 	}
 	obj.mutex.Unlock()
+	obj.wg.Add(1)
 	go func() {
+		defer obj.wg.Done()
 		for {
 			select {
 			case _, ok := <-obj.timer: // reset signal channel
@@ -359,24 +362,25 @@ func (obj *convergerUID) StartTimer() (func() error, error) {
 }
 
 // ResetTimer resets the counter to zero if using a StartTimer internally.
-func (obj *convergerUID) ResetTimer() error {
+func (obj *cuid) ResetTimer() error {
 	obj.mutex.Lock()
 	defer obj.mutex.Unlock()
 	if obj.running {
 		obj.timer <- struct{}{} // send the reset message
 		return nil
 	}
-	return fmt.Errorf("Timer hasn't been started!")
+	return fmt.Errorf("timer hasn't been started")
 }
 
 // StopTimer stops the running timer permanently until a StartTimer is run.
-func (obj *convergerUID) StopTimer() error {
+func (obj *cuid) StopTimer() error {
 	obj.mutex.Lock()
 	defer obj.mutex.Unlock()
 	if !obj.running {
-		return fmt.Errorf("Timer isn't running!")
+		return fmt.Errorf("timer isn't running")
 	}
 	close(obj.timer)
+	obj.wg.Wait()
 	obj.running = false
 	return nil
 }

doc.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify

docs/.gitignore

@@ -0,0 +1,2 @@
+mgmt-documentation.pdf
+_build

docs/Makefile

@@ -0,0 +1,20 @@
+# Minimal makefile for Sphinx documentation
+#
+
+# You can set these variables from the command line.
+SPHINXOPTS    =
+SPHINXBUILD   = sphinx-build
+SPHINXPROJ    = mgmt
+SOURCEDIR     = .
+BUILDDIR      = _build
+
+# Put it first so that "make" without argument is like "make help".
+help:
+	@$(SPHINXBUILD) -M help "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)
+
+.PHONY: help Makefile
+
+# Catch-all target: route all unknown targets to Sphinx using the new
+# "make mode" option.  $(O) is meant as a shortcut for $(SPHINXOPTS).
+%: Makefile
+	@$(SPHINXBUILD) -M $@ "$(SOURCEDIR)" "$(BUILDDIR)" $(SPHINXOPTS) $(O)

docs/conf.py

@@ -0,0 +1,158 @@
+# -*- coding: utf-8 -*-
+#
+# mgmt documentation build configuration file, created by
+# sphinx-quickstart on Wed Feb 15 21:34:09 2017.
+#
+# This file is execfile()d with the current directory set to its
+# containing dir.
+#
+# Note that not all possible configuration values are present in this
+# autogenerated file.
+#
+# All configuration values have a default; values that are commented out
+# serve to show the default.
+
+# If extensions (or modules to document with autodoc) are in another directory,
+# add these directories to sys.path here. If the directory is relative to the
+# documentation root, use os.path.abspath to make it absolute, like shown here.
+#
+# import os
+# import sys
+# sys.path.insert(0, os.path.abspath('.'))
+
+from recommonmark.parser import CommonMarkParser
+
+# -- General configuration ------------------------------------------------
+
+# If your documentation needs a minimal Sphinx version, state it here.
+#
+# needs_sphinx = '1.0'
+
+# Add any Sphinx extension module names here, as strings. They can be
+# extensions coming with Sphinx (named 'sphinx.ext.*') or your custom
+# ones.
+extensions = []
+
+# Add any paths that contain templates here, relative to this directory.
+templates_path = ['_templates']
+
+# The suffix(es) of source filenames.
+# You can specify multiple suffix as a list of string:
+#
+
+source_parsers = {
+        '.md': CommonMarkParser,
+}
+
+source_suffix = ['.rst', '.md']
+
+# The master toctree document.
+master_doc = 'index'
+
+# General information about the project.
+project = u'mgmt'
+copyright = u'2013-2017+ James Shubin and the project contributors'
+author = u'James Shubin'
+
+# The version info for the project you're documenting, acts as replacement for
+# |version| and |release|, also used in various other places throughout the
+# built documents.
+#
+# The short X.Y version.
+version = u''
+# The full version, including alpha/beta/rc tags.
+release = u''
+
+# The language for content autogenerated by Sphinx. Refer to documentation
+# for a list of supported languages.
+#
+# This is also used if you do content translation via gettext catalogs.
+# Usually you set "language" from the command line for these cases.
+language = None
+
+# List of patterns, relative to source directory, that match files and
+# directories to ignore when looking for source files.
+# This patterns also effect to html_static_path and html_extra_path
+exclude_patterns = ['_build', 'Thumbs.db', '.DS_Store', 'venv']
+
+# The name of the Pygments (syntax highlighting) style to use.
+pygments_style = 'sphinx'
+
+# If true, `todo` and `todoList` produce output, else they produce nothing.
+todo_include_todos = False
+
+
+# -- Options for HTML output ----------------------------------------------
+
+# The theme to use for HTML and HTML Help pages.  See the documentation for
+# a list of builtin themes.
+#
+#html_theme = 'alabaster'
+
+# Theme options are theme-specific and customize the look and feel of a theme
+# further.  For a list of options available for each theme, see the
+# documentation.
+#
+# html_theme_options = {}
+
+# Add any paths that contain custom static files (such as style sheets) here,
+# relative to this directory. They are copied after the builtin static files,
+# so a file named "default.css" will overwrite the builtin "default.css".
+html_static_path = ['_static']
+
+
+# -- Options for HTMLHelp output ------------------------------------------
+
+# Output file base name for HTML help builder.
+htmlhelp_basename = 'mgmtdoc'
+
+
+# -- Options for LaTeX output ---------------------------------------------
+
+latex_elements = {
+    # The paper size ('letterpaper' or 'a4paper').
+    #
+    # 'papersize': 'letterpaper',
+
+    # The font size ('10pt', '11pt' or '12pt').
+    #
+    # 'pointsize': '10pt',
+
+    # Additional stuff for the LaTeX preamble.
+    #
+    # 'preamble': '',
+
+    # Latex figure (float) alignment
+    #
+    # 'figure_align': 'htbp',
+}
+
+# Grouping the document tree into LaTeX files. List of tuples
+# (source start file, target name, title,
+#  author, documentclass [howto, manual, or own class]).
+latex_documents = [
+    (master_doc, 'mgmt.tex', u'mgmt Documentation',
+     u'James Shubin', 'manual'),
+]
+
+
+# -- Options for manual page output ---------------------------------------
+
+# One entry per manual page. List of tuples
+# (source start file, name, description, authors, manual section).
+man_pages = [
+    (master_doc, 'mgmt', u'mgmt Documentation',
+     [author], 1)
+]
+
+
+# -- Options for Texinfo output -------------------------------------------
+
+# Grouping the document tree into Texinfo files. List of tuples
+# (source start file, target name, title, author,
+#  dir menu entry, description, category)
+texinfo_documents = [
+    (master_doc, 'mgmt', u'mgmt Documentation',
+     author, 'mgmt', 'A next generation config management prototype!',
+     'Miscellaneous'),
+]

docs/documentation.md

@@ -1,57 +1,16 @@
-#mgmt
+# mgmt
 
-<!--
-Mgmt
-Copyright (C) 2013-2016+ James Shubin and the project contributors
-Written by James Shubin <james@shubin.ca> and the project contributors
+Available from:
+[https://github.com/purpleidea/mgmt/](https://github.com/purpleidea/mgmt/)
 
-This program is free software: you can redistribute it and/or modify
-it under the terms of the GNU Affero General Public License as published by
-the Free Software Foundation, either version 3 of the License, or
-(at your option) any later version.
+This documentation is available in: [Markdown](https://github.com/purpleidea/mgmt/blob/master/docs/documentation.md) or [PDF](https://pdfdoc-purpleidea.rhcloud.com/pdf/https://github.com/purpleidea/mgmt/blob/master/docs/documentation.md) format.
 
-This program is distributed in the hope that it will be useful,
-but WITHOUT ANY WARRANTY; without even the implied warranty of
-MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
-GNU Affero General Public License for more details.
-
-You should have received a copy of the GNU Affero General Public License
-along with this program.  If not, see <http://www.gnu.org/licenses/>.
--->
-
-##mgmt by [James](https://ttboj.wordpress.com/)
-####Available from:
-####[https://github.com/purpleidea/mgmt/](https://github.com/purpleidea/mgmt/)
-
-####This documentation is available in: [Markdown](https://github.com/purpleidea/mgmt/blob/master/DOCUMENTATION.md) or [PDF](https://pdfdoc-purpleidea.rhcloud.com/pdf/https://github.com/purpleidea/mgmt/blob/master/DOCUMENTATION.md) format.
-
-####Table of Contents
-
-1. [Overview](#overview)
-2. [Project description - What the project does](#project-description)
-3. [Setup - Getting started with mgmt](#setup)
-4. [Features - All things mgmt can do](#features)
-	* [Autoedges - Automatic resource relationships](#autoedges)
-	* [Autogrouping - Automatic resource grouping](#autogrouping)
-	* [Automatic clustering - Automatic cluster management](#automatic-clustering)
-	* [Remote mode - Remote "agent-less" execution](#remote-agent-less-mode)
-	* [Puppet support - write manifest code for mgmt](#puppet-support)
-5. [Resources - All built-in primitives](#resources)
-6. [Usage/FAQ - Notes on usage and frequently asked questions](#usage-and-frequently-asked-questions)
-7. [Reference - Detailed reference](#reference)
-	* [Meta parameters](#meta-parameters)
-	* [Graph definition file](#graph-definition-file)
-	* [Command line](#command-line)
-8. [Examples - Example configurations](#examples)
-9. [Development - Background on module development and reporting bugs](#development)
-10. [Authors - Authors and contact information](#authors)
-
-##Overview
+## Overview
 
 The `mgmt` tool is a next generation config management prototype. It's not yet
 ready for production, but we hope to get there soon. Get involved today!
 
-##Project Description
+## Project Description
 
 The mgmt tool is a distributed, event driven, config management tool, that
 supports parallel execution, and librarification to be used as the management
@@ -63,11 +22,14 @@ For more information, you may like to read some blog posts from the author:
 * [Automatic edges in mgmt](https://ttboj.wordpress.com/2016/03/14/automatic-edges-in-mgmt/)
 * [Automatic grouping in mgmt](https://ttboj.wordpress.com/2016/03/30/automatic-grouping-in-mgmt/)
 * [Automatic clustering in mgmt](https://ttboj.wordpress.com/2016/06/20/automatic-clustering-in-mgmt/)
+* [Remote execution in mgmt](https://ttboj.wordpress.com/2016/10/07/remote-execution-in-mgmt/)
+* [Send/Recv in mgmt](https://ttboj.wordpress.com/2016/12/07/sendrecv-in-mgmt/)
+* [Metaparameters in mgmt](https://ttboj.wordpress.com/2017/03/01/metaparameters-in-mgmt/)
 
 There is also an [introductory video](http://meetings-archive.debian.net/pub/debian-meetings/2016/debconf16/Next_Generation_Config_Mgmt.webm) available.
 Older videos and other material [is available](https://github.com/purpleidea/mgmt/#on-the-web).
 
-##Setup
+## Setup
 
 During this prototype phase, the tool can be run out of the source directory.
 You'll probably want to use ```./run.sh run --yaml examples/graph1.yaml``` to
@@ -75,12 +37,12 @@ get started. Beware that this _can_ cause data loss. Understand what you're
 doing first, or perform these actions in a virtual environment such as the one
 provided by [Oh-My-Vagrant](https://github.com/purpleidea/oh-my-vagrant).
 
-##Features
+## Features
 
 This section details the numerous features of mgmt and some caveats you might
 need to be aware of.
 
-###Autoedges
+### Autoedges
 
 Automatic edges, or AutoEdges, is the mechanism in mgmt by which it will
 automatically create dependencies for you between resources. For example,
@@ -89,7 +51,7 @@ automatically ensure that any file resource you declare that matches a
 file installed by your package resource will only be processed after the
 package is installed.
 
-####Controlling autoedges
+#### Controlling autoedges
 
 Though autoedges is likely to be very helpful and avoid you having to declare
 all dependencies explicitly, there are cases where this behaviour is
@@ -106,12 +68,12 @@ installation of the `mysql-server` package.
 You can disable autoedges for a resource by setting the `autoedge` key on
 the meta attributes of that resource to `false`.
 
-####Blog post
+#### Blog post
 
 You can read the introductory blog post about this topic here:
 [https://ttboj.wordpress.com/2016/03/14/automatic-edges-in-mgmt/](https://ttboj.wordpress.com/2016/03/14/automatic-edges-in-mgmt/)
 
-###Autogrouping
+### Autogrouping
 
 Automatic grouping or AutoGroup is the mechanism in mgmt by which it will
 automatically group multiple resource vertices into a single one. This is
@@ -125,12 +87,12 @@ used for other use cases too.
 You can disable autogrouping for a resource by setting the `autogroup` key on
 the meta attributes of that resource to `false`.
 
-####Blog post
+#### Blog post
 
 You can read the introductory blog post about this topic here:
 [https://ttboj.wordpress.com/2016/03/30/automatic-grouping-in-mgmt/](https://ttboj.wordpress.com/2016/03/30/automatic-grouping-in-mgmt/)
 
-###Automatic clustering
+### Automatic clustering
 
 Automatic clustering is a feature by which mgmt automatically builds, scales,
 and manages the embedded etcd cluster which is compiled into mgmt itself. It is
@@ -141,12 +103,12 @@ If you prefer to avoid this feature. you can always opt to use an existing etcd
 cluster that is managed separately from mgmt by pointing your mgmt agents at it
 with the `--seeds` variable.
 
-####Blog post
+#### Blog post
 
 You can read the introductory blog post about this topic here:
 [https://ttboj.wordpress.com/2016/06/20/automatic-clustering-in-mgmt/](https://ttboj.wordpress.com/2016/06/20/automatic-clustering-in-mgmt/)
 
-###Remote ("agent-less") mode
+### Remote ("agent-less") mode
 
 Remote mode is a special mode that lets you kick off mgmt runs on one or more
 remote machines which are only accessible via SSH. In this mode the initiating
@@ -168,12 +130,12 @@ entire set of running mgmt agents will need to all simultaneously converge for
 the group to exit. This is particularly useful for bootstrapping new clusters
 which need to exchange information that is only available at run time.
 
-####Blog post
+#### Blog post
 
 You can read the introductory blog post about this topic here:
 [https://ttboj.wordpress.com/2016/10/07/remote-execution-in-mgmt/](https://ttboj.wordpress.com/2016/10/07/remote-execution-in-mgmt/)
 
-###Puppet support
+### Puppet support
 
 You can supply a Puppet manifest instead of creating the (YAML) graph manually.
 Puppet must be installed and in `mgmt`'s search path. You also need the
@@ -195,12 +157,12 @@ Invoke `mgmt` with the `--puppet` switch, which supports 3 variants:
 
 For more details and caveats see [Puppet.md](Puppet.md).
 
-####Blog post
+#### Blog post
 
 An introductory post on the Puppet support is on
 [Felix's blog](http://ffrank.github.io/features/2016/06/19/puppet-powered-mgmt/).
 
-##Resources
+## Resources
 
 This section lists all the built-in resources and their properties. The
 resource primitives in `mgmt` are typically more powerful than resources in
@@ -216,9 +178,11 @@ meta parameters aren't very useful when combined with certain resources, but
 in general, it should be fairly obvious, such as when combining the `noop` meta
 parameter with the [Noop](#Noop) resource.
 
+* [Augeas](#Augeas): Manipulate files using augeas.
 * [Exec](#Exec): Execute shell commands on the system.
 * [File](#File): Manage files and directories.
 * [Hostname](#Hostname): Manages the hostname on the system.
+* [KV](#KV): Set a key value pair in our shared world database.
 * [Msg](#Msg): Send log messages.
 * [Noop](#Noop): A simple resource that does nothing.
 * [Nspawn](#Nspawn): Manage systemd-machined nspawn containers.
@@ -228,45 +192,60 @@ parameter with the [Noop](#Noop) resource.
 * [Timer](#Timer): Manage system systemd services.
 * [Virt](#Virt): Manage virtual machines with libvirt.
 
-###Exec
+
+### Augeas
+
+The augeas resource uses [augeas](http://augeas.net/) commands to manipulate
+files.
+
+### Exec
 
 The exec resource can execute commands on your system.
 
-###File
+### File
 
 The file resource manages files and directories. In `mgmt`, directories are
 identified by a trailing slash in their path name. File have no such slash.
 
-####Path
+It has the following properties:
+
+- `path`: file path (directories have a trailing slash here)
+- `content`: raw file content
+- `state`: either `exists` (the default value) or `absent`
+- `mode`: octal unix file permissions
+- `owner`: username or uid for the file owner
+- `group`: group name or gid for the file group
+
+#### Path
 
 The path property specifies the file or directory that we are managing.
 
-####Content
+#### Content
 
 The content property is a string that specifies the desired file contents.
 
-####Source
+#### Source
 
 The source property points to a source file or directory path that we wish to
 copy over and use as the desired contents for our resource.
 
-####State
+#### State
 
 The state property describes the action we'd like to apply for the resource. The
 possible values are: `exists` and `absent`.
 
-####Recurse
+#### Recurse
 
 The recurse property limits whether file resource operations should recurse into
 and monitor directory contents with a depth greater than one.
 
-####Force
+#### Force
 
 The force property is required if we want the file resource to be able to change
 a file into a directory or vice-versa. If such a change is needed, but the force
 property is not set to `true`, then this file resource will error.
 
-###Hostname
+### Hostname
 
 The hostname resource manages static, transient/dynamic and pretty hostnames
 on the system and watches them for changes.
@@ -290,55 +269,79 @@ The pretty hostname is a free-form UTF8 host name for presentation to the user.
 Hostname is the fallback value for all 3 fields above, if only `hostname` is
 specified, it will set all 3 fields to this value.
 
-###Msg
+### KV
+
+The KV resource sets a key and value pair in the global world database. This is
+quite useful for setting a flag after a number of resources have run. It will
+ignore database updates to the value that are greater in compare order than the
+requested key if the `SkipLessThan` parameter is set to true. If we receive a
+refresh, then the stored value will be reset to the requested value even if the
+stored value is greater.
+
+#### Key
+The string key used to store the key.
+
+#### Value
+The string value to set. This can also be set via Send/Recv.
+
+#### SkipLessThan
+If this parameter is set to `true`, then it will ignore updating the value as
+long as the database versions are greater than the requested value. The compare
+operation used is based on the `SkipCmpStyle` parameter.
+
+#### SkipCmpStyle
+By default this converts the string values to integers and compares them as you
+would expect.
+
+### Msg
 
 The msg resource sends messages to the main log, or an external service such
 as systemd's journal.
 
-###Noop
+### Noop
 
 The noop resource does absolutely nothing. It does have some utility in testing
 `mgmt` and also as a placeholder in the resource graph.
 
-###Nspawn
+### Nspawn
 
 The nspawn resource is used to manage systemd-machined style containers.
 
-###Password
+### Password
 
 The password resource can generate a random string to be used as a password. It
 will re-generate the password if it receives a refresh notification.
 
-###Pkg
+### Pkg
 
 The pkg resource is used to manage system packages. This resource works on many
 different distributions because it uses the underlying packagekit facility which
 supports different backends for different environments. This ensures that we
 have great Debian (deb/dpkg) and Fedora (rpm/dnf) support simultaneously.
 
-###Svc
+### Svc
 
 The service resource is still very WIP. Please help us my improving it!
 
-###Timer
+### Timer
 
 This resource needs better documentation. Please help us my improving it!
 
-###Virt
+### Virt
 
 The virt resource can manage virtual machines via libvirt.
 
-##Usage and frequently asked questions
+## Usage and frequently asked questions
 (Send your questions as a patch to this FAQ! I'll review it, merge it, and
 respond by commit with the answer.)
 
-###Why did you start this project?
+### Why did you start this project?
 
 I wanted a next generation config management solution that didn't have all of
 the design flaws or limitations that the current generation of tools do, and no
 tool existed!
 
-###Why did you use etcd? What about consul?
+### Why did you use etcd? What about consul?
 
 Etcd and consul are both written in golang, which made them the top two
 contenders for my prototype. Ultimately a choice had to be made, and etcd was
@@ -346,7 +349,7 @@ chosen, but it was also somewhat arbitrary. If there is available interest,
 good reasoning, *and* patches, then we would consider either switching or
 supporting both, but this is not a high priority at this time.
 
-###Can I use an existing etcd cluster instead of the automatic embedded servers?
+### Can I use an existing etcd cluster instead of the automatic embedded servers?
 
 Yes, it's possible to use an existing etcd cluster instead of the automatic,
 elastic embedded etcd servers. To do so, simply point to the cluster with the
@@ -357,7 +360,7 @@ The downside to this approach is that you won't benefit from the automatic
 elastic nature of the embedded etcd servers, and that you're responsible if you
 accidentally break your etcd cluster, or if you use an unsupported version.
 
-###What does the error message about an inconsistent dataDir mean?
+### What does the error message about an inconsistent dataDir mean?
 
 If you get an error message similar to:
 
@@ -375,7 +378,7 @@ starting up, and as a result, a default endpoint never gets added. The solution
 is to either reconcile the mistake, and if there is no important data saved, you
 can remove the etcd dataDir. This is typically `/var/lib/mgmt/etcd/member/`.
 
-###Why do resources have both a `Compare` method and an `IFF` (on the UID) method?
+### Why do resources have both a `Compare` method and an `IFF` (on the UID) method?
 
 The `Compare()` methods are for determining if two resources are effectively the
 same, which is used to make graph change delta's efficient. This is when we want
@@ -391,14 +394,14 @@ equality. In the future it might be helpful or sane to merge the two similar
 comparison functions although for now they are separate because they are
 actually answer different questions.
 
-###Did you know that there is a band named `MGMT`?
+### Did you know that there is a band named `MGMT`?
 
 I didn't realize this when naming the project, and it is accidental. After much
 anguishing, I chose the name because it was short and I thought it was
 appropriately descriptive. If you need a less ambiguous search term or phrase,
 you can try using `mgmtconfig` or `mgmt config`.
 
-###You didn't answer my question, or I have a question!
+### You didn't answer my question, or I have a question!
 
 It's best to ask on [IRC](https://webchat.freenode.net/?channels=#mgmtconfig)
 to see if someone can help you. Once we get a big enough community going, we'll
@@ -408,40 +411,41 @@ and I'll do my best to help. If you have a good question, please add it as a
 patch to this documentation. I'll merge your question, and add a patch with the
 answer!
 
-##Reference
+## Reference
 Please note that there are a number of undocumented options. For more
 information on these options, please view the source at:
 [https://github.com/purpleidea/mgmt/](https://github.com/purpleidea/mgmt/).
 If you feel that a well used option needs documenting here, please patch it!
 
-###Overview of reference
+### Overview of reference
 * [Meta parameters](#meta-parameters): List of available resource meta parameters.
 * [Graph definition file](#graph-definition-file): Main graph definition file.
 * [Command line](#command-line): Command line parameters.
+* [Compilation options](#compilation-options): Compilation options.
 
-###Meta parameters
+### Meta parameters
 These meta parameters are special parameters (or properties) which can apply to
 any resource. The usefulness of doing so will depend on the particular meta
 parameter and resource combination.
 
-####AutoEdge
+#### AutoEdge
 Boolean. Should we generate auto edges for this resource?
 
-####AutoGroup
+#### AutoGroup
 Boolean. Should we attempt to automatically group this resource with others?
 
-####Noop
+#### Noop
 Boolean. Should the Apply portion of the CheckApply method of the resource
 make any changes? Noop is a concatenation of no-operation.
 
-####Retry
+#### Retry
 Integer. The number of times to retry running the resource on error. Use -1 for
 infinite. This currently applies for both the Watch operation (which can fail)
 and for the CheckApply operation. While they could have separate values, I've
 decided to use the same ones for both until there's a proper reason to want to
 do something differently for the Watch errors.
 
-####Delay
+#### Delay
 Integer. Number of milliseconds to wait between retries. The same value is
 shared between the Watch and CheckApply retries. This currently applies for both
 the Watch operation (which can fail) and for the CheckApply operation. While
@@ -449,63 +453,113 @@ they could have separate values, I've decided to use the same ones for both
 until there's a proper reason to want to do something differently for the Watch
 errors.
 
-###Graph definition file
+#### Poll
+Integer. Number of seconds to wait between `CheckApply` checks. If this is
+greater than zero, then the standard event based `Watch` mechanism for this
+resource is replaced with a simple polling mechanism. In general, this is not
+recommended, unless you have a very good reason for doing so.
+
+Please keep in mind that if you have a resource which changes every `I` seconds,
+and you poll it every `J` seconds, and you've asked for a converged timeout of
+`K` seconds, and `I <= J <= K`, then your graph will likely never converge.
+
+When polling, the system detects that a resource is not converged if its
+`CheckApply` method returns false. This allows a resource which changes every
+`I` seconds, and which is polled every `J` seconds, and with a converged timeout
+of `K` seconds to still converge when `J <= K`, as long as `I > J || I > K`,
+which is another way of saying that if the resource finally settles down to give
+the graph enough time, it can probably converge.
+
+#### Limit
+Float. Maximum rate of `CheckApply` runs started per second. Useful to limit
+an especially _eventful_ process from causing excessive checks to run. This
+defaults to `+Infinity` which adds no limiting. If you change this value, you
+will also need to change the `Burst` value to a non-zero value. Please see the
+[rate](https://godoc.org/golang.org/x/time/rate) package for more information.
+
+#### Burst
+Integer. Burst is the maximum number of runs which can happen without invoking
+the rate limiter as designated by the `Limit` value. If the `Limit` is not set
+to `+Infinity`, this must be a non-zero value. Please see the
+[rate](https://godoc.org/golang.org/x/time/rate) package for more information.
+
+#### Sema
+List of string ids. Sema is a P/V style counting semaphore which can be used to
+limit parallelism during the CheckApply phase of resource execution. Each
+resource can have `N` different semaphores which share a graph global namespace.
+Each semaphore has a maximum count associated with it. The default value of the
+size is 1 (one) if size is unspecified. Each string id is the unique id of the
+semaphore. If the id contains a trailing colon (:) followed by a positive
+integer, then that value is the max size for that semaphore. Valid semaphore
+id's include: `some_id`, `hello:42`, `not:smart:4` and `:13`. It is expected
+that the last bare example be only used by the engine to add a global semaphore.
+
+### Graph definition file
 graph.yaml is the compiled graph definition file. The format is currently
 undocumented, but by looking through the [examples/](https://github.com/purpleidea/mgmt/tree/master/examples)
 you can probably figure out most of it, as it's fairly intuitive.
 
-###Command line
+### Command line
 The main interface to the `mgmt` tool is the command line. For the most recent
 documentation, please run `mgmt --help`.
 
-####`--yaml <graph.yaml>`
+#### `--yaml <graph.yaml>`
 Point to a graph file to run.
 
-####`--converged-timeout <seconds>`
+#### `--converged-timeout <seconds>`
 Exit if the machine has converged for approximately this many seconds.
 
-####`--max-runtime <seconds>`
+#### `--max-runtime <seconds>`
 Exit when the agent has run for approximately this many seconds. This is not
 generally recommended, but may be useful for users who know what they're doing.
 
-####`--noop`
+#### `--noop`
 Globally force all resources into no-op mode. This also disables the export to
 etcd functionality, but does not disable resource collection, however all
 resources that are collected will have their individual noop settings set.
 
-####`--remote <graph.yaml>`
+#### `--sema <size>`
+Globally add a counting semaphore of this size to each resource in the graph.
+The semaphore will get given an id of `:size`. In other words if you specify a
+size of 42, you can expect a semaphore if named: `:42`. It is expected that
+consumers of the semaphore metaparameter always include a prefix to avoid a
+collision with this globally defined semaphore. The size value must be greater
+than zero at this time. The traditional non-parallel execution found in config
+management tools such as `Puppet` can be obtained with `--sema 1`.
+
+#### `--remote <graph.yaml>`
 Point to a graph file to run on the remote host specified within. This parameter
 can be used multiple times if you'd like to remotely run on multiple hosts in
 parallel.
 
-####`--allow-interactive`
+#### `--allow-interactive`
 Allow interactive prompting for SSH passwords if there is no authentication
 method that works.
 
-####`--ssh-priv-id-rsa`
+#### `--ssh-priv-id-rsa`
 Specify the path for finding SSH keys. This defaults to `~/.ssh/id_rsa`. To
 never use this method of authentication, set this to the empty string.
 
-####`--cconns`
+#### `--cconns`
 The maximum number of concurrent remote ssh connections to run. This defaults
 to `0`, which means unlimited.
 
-####`--no-caching`
+#### `--no-caching`
 Don't allow remote caching of the remote execution binary. This will require
 the binary to be copied over for every remote execution, but it limits the
 likelihood that there is leftover information from the configuration process.
 
-####`--prefix <path>`
+#### `--prefix <path>`
 Specify a path to a custom working directory prefix. This directory will get
 created if it does not exist. This usually defaults to `/var/lib/mgmt/`. This
 can't be combined with the `--tmp-prefix` option. It can be combined with the
 `--allow-tmp-prefix` option.
 
-####`--tmp-prefix`
+#### `--tmp-prefix`
 If this option is specified, a temporary prefix will be used instead of the
 default prefix. This can't be combined with the `--prefix` option.
 
-####`--allow-tmp-prefix`
+#### `--allow-tmp-prefix`
 If this option is specified, we will attempt to fall back to a temporary prefix
 if the primary prefix couldn't be created. This is useful for avoiding failures
 in environments where the primary prefix may or may not be available, but you'd
@@ -513,7 +567,35 @@ like to try. The canonical example is when running `mgmt` with `--remote` there
 might be a cached copy of the binary in the primary prefix, but in case there's
 no binary available continue working in a temporary directory to avoid failure.
 
-##Examples
+### Compilation options
+
+You can control some compilation variables by using environment variables.
+
+#### Disable libvirt support
+
+If you wish to compile mgmt without libvirt, you can use the following command:
+
+```
+GOTAGS=novirt make build
+```
+
+#### Disable augeas support
+
+If you wish to compile mgmt without augeas support, you can use the following command:
+
+```
+GOTAGS=noaugeas make build
+```
+
+#### Combining compile-time flags
+
+You can combine multiple tags by using a space-separated list:
+
+```
+GOTAGS="noaugeas novirt" make build
+```
+
+## Examples
 For example configurations, please consult the [examples/](https://github.com/purpleidea/mgmt/tree/master/examples) directory in the git
 source repository. It is available from:
 
@@ -541,7 +623,7 @@ EOF
 sudo systemctl daemon-reload
 ```
 
-##Development
+## Development
 
 This is a project that I started in my free time in 2013. Development is driven
 by all of our collective patches! Dive right in, and start hacking!
@@ -551,9 +633,9 @@ You can follow along [on my technical blog](https://ttboj.wordpress.com/).
 
 To report any bugs, please file a ticket at: [https://github.com/purpleidea/mgmt/issues](https://github.com/purpleidea/mgmt/issues).
 
-##Authors
+## Authors
 
-Copyright (C) 2013-2016+ James Shubin and the project contributors
+Copyright (C) 2013-2017+ James Shubin and the project contributors
 
 Please see the
 [AUTHORS](https://github.com/purpleidea/mgmt/tree/master/AUTHORS) file

docs/index.rst

@@ -0,0 +1,17 @@
+.. mgmt documentation master file, created by
+   sphinx-quickstart on Wed Feb 15 21:34:09 2017.
+   You can adapt this file completely to your liking, but it should at least
+   contain the root `toctree` directive.
+
+Welcome to mgmt's documentation!
+================================
+
+.. toctree::
+   :maxdepth: 2
+   :caption: Contents:
+
+   documentation
+   quick-start-guide
+   resource-guide
+   prometheus
+   puppet-guide

docs/prometheus.md

@@ -0,0 +1,66 @@
+# Prometheus support
+
+Mgmt comes with a built-in prometheus support. It is disabled by default, and
+can be enabled with the `--prometheus` command line switch.
+
+By default, the prometheus instance will listen on [`127.0.0.1:9233`][pd]. You
+can change this setting by using the `--prometheus-listen` cli option:
+
+To have mgmt prometheus bind interface on 0.0.0.0:45001, use:
+`./mgmt r --prometheus --prometheus-listen :45001`
+
+## Metrics
+
+Mgmt exposes three kinds of resources: _go_ metrics, _etcd_ metrics and _mgmt_
+metrics.
+
+### go metrics
+
+We use the [prometheus go_collector][pgc] to expose go metrics. Those metrics
+are mainly useful for debugging and perf testing.
+
+### etcd metrics
+
+mgmt exposes etcd metrics. Read more in the [upstream documentation][etcdm]
+
+### mgmt metrics
+
+Here is a list of the metrics we provide:
+
+- `mgmt_resources_total`: The number of resources that mgmt is managing
+- `mgmt_checkapply_total`: The number of CheckApply's that mgmt has run
+- `mgmt_failures_total`: The number of resources that have failed
+- `mgmt_failures_current`: The number of resources that have failed
+- `mgmt_graph_start_time_seconds`: Start time of the current graph since unix epoch in seconds
+
+For each metric, you will get some extra labels:
+
+- `kind`: The kind of mgmt resource
+
+For `mgmt_checkapply_total`, those extra labels are set:
+
+- `eventful`: "true" or "false", if the CheckApply triggered some changes
+- `errorful`: "true" or "false", if the CheckApply reported an error
+- `apply`: "true" or "false", if the CheckApply ran in apply or noop mode
+
+## Alerting
+
+You can use prometheus to alert you upon changes or failures. We do not provide
+such templates yet, but we plan to provide some examples in this repository.
+Patches welcome!
+
+## Grafana
+
+We do not have grafana dashboards yet. Patches welcome!
+
+## External resources
+
+- [prometheus website](https://prometheus.io/)
+- [prometheus documentation](https://prometheus.io/docs/introduction/overview/)
+- [prometheus best practices regarding metrics
+  naming](https://prometheus.io/docs/practices/naming/)
+- [grafana website](http://grafana.org/)
+
+[pgc]: https://github.com/prometheus/client_golang/blob/master/prometheus/go_collector.go
+[etcdm]: https://coreos.com/etcd/docs/latest/metrics.html
+[pd]: https://github.com/prometheus/prometheus/wiki/Default-port-allocation

docs/puppet-guide.md

@@ -1,22 +1,13 @@
-#mgmt Puppet support
-
-1. [Prerequisites](#prerequisites)
-    * [Testing the Puppet side](#testing-the-puppet-side)
-2. [Writing a suitable manifest](#writing-a-suitable-manifest)
-    * [Unsupported attributes](#unsupported-attributes)
-    * [Unsupported resources](#unsupported-resources)
-    * [Avoiding common warnings](#avoiding-common-warnings)
-3. [Configuring Puppet](#configuring-puppet)
-4. [Caveats](#caveats)
+# Puppet guide
 
 `mgmt` can use Puppet as its source for the configuration graph.
 This document goes into detail on how this works, and lists
 some pitfalls and limitations.
 
 For basic instructions on how to use the Puppet support, see
-the [main documentation](DOCUMENTATION.md#puppet-support).
+the [main documentation](documentation.md#puppet-support).
 
-##Prerequisites
+## Prerequisites
 
 You need Puppet installed in your system. It is not important how you
 get it. On the most common Linux distributions, you can use packages
@@ -29,14 +20,16 @@ Any release of Puppet's 3.x and 4.x series should be suitable for use with
 `mgmt`. Most importantly, make sure to install the `ffrank-mgmtgraph` Puppet
 module (referred to below as "the translator module").
 
-    puppet module install ffrank-mgmtgraph
+```
+puppet module install ffrank-mgmtgraph
+```
 
 Please note that the module is not required on your Puppet master (if you
 use a master/agent setup). It's needed on the machine that runs `mgmt`.
 You can install the module on the master anyway, so that it gets distributed
 to your agents through Puppet's `pluginsync` mechanism.
 
-###Testing the Puppet side
+### Testing the Puppet side
 
 The following command should run successfully and print a YAML hash on your
 terminal:
@@ -48,9 +41,9 @@ puppet mgmtgraph print --code 'file { "/tmp/mgmt-test": ensure => present }'
 You can use this CLI to test any manifests before handing them straight
 to `mgmt`.
 
-##Writing a suitable manifest
+## Writing a suitable manifest
 
-###Unsupported attributes
+### Unsupported attributes
 
 `mgmt` inherited its resource module from Puppet, so by and large, it's quite
 possible to express `mgmt` graphs in terms of Puppet manifests. However,
@@ -62,8 +55,10 @@ For example, at the time of writing this, the `file` type in `mgmt` had no
 notion of permissions (the file `mode`) yet. This lead to the following
 warning (among others that will be discussed below):
 
-    $ puppet mgmtgraph print --code 'file { "/tmp/foo": mode => "0600" }'
-    Warning: cannot translate: File[/tmp/foo] { mode => "600" } (attribute is ignored)
+```
+$ puppet mgmtgraph print --code 'file { "/tmp/foo": mode => "0600" }'
+Warning: cannot translate: File[/tmp/foo] { mode => "600" } (attribute is ignored)
+```
 
 This is a heads-up for the user, because the resulting `mgmt` graph will
 in fact not pass this information to the `/tmp/foo` file resource, and
@@ -71,7 +66,7 @@ in fact not pass this information to the `/tmp/foo` file resource, and
 manifests that are written expressly for `mgmt` is not sensible and should
 be avoided.
 
-###Unsupported resources
+### Unsupported resources
 
 Puppet has a fairly large number of
 [built-in types](https://docs.puppet.com/puppet/latest/reference/type.html),
@@ -91,28 +86,32 @@ this overhead can amount to several orders of magnitude.
 
 Avoid Puppet types that `mgmt` does not implement (yet).
 
-###Avoiding common warnings
+### Avoiding common warnings
 
 Many resource parameters in Puppet take default values. For the most part,
 the translator module just ignores them. However, there are cases in which
 Puppet will default to convenient behavior that `mgmt` cannot quite replicate.
 For example, translating a plain `file` resource will lead to a warning message:
 
-    $ puppet mgmtgraph print --code 'file { "/tmp/mgmt-test": }'
-    Warning: File[/tmp/mgmt-test] uses the 'puppet' file bucket, which mgmt cannot do. There will be no backup copies!
+```
+$ puppet mgmtgraph print --code 'file { "/tmp/mgmt-test": }'
+Warning: File[/tmp/mgmt-test] uses the 'puppet' file bucket, which mgmt cannot do. There will be no backup copies!
+```
 
 The reason is that per default, Puppet assumes the following parameter value
 (among others)
 
 ```puppet
 file { "/tmp/mgmt-test":
-  backup => 'puppet',
+	backup => 'puppet',
 }
 ```
 
 To avoid this, specify the parameter explicitly:
 
-    $ puppet mgmtgraph print --code 'file { "/tmp/mgmt-test": backup => false }'
+```
+$ puppet mgmtgraph print --code 'file { "/tmp/mgmt-test": backup => false }'
+```
 
 This is tedious in a more complex manifest. A good simplification is the
 following [resource default](https://docs.puppet.com/puppet/latest/reference/lang_defaults.html)
@@ -125,7 +124,7 @@ File { backup => false }
 If you encounter similar warnings from other types and/or parameters,
 use the same approach to silence them if possible.
 
-##Configuring Puppet
+## Configuring Puppet
 
 Since `mgmt` uses an actual Puppet CLI behind the scenes, you might
 need to tweak some of Puppet's runtime options in order to make it
@@ -143,16 +142,20 @@ control all of them, through its `--puppet-conf` option. It allows
 you to specify which `puppet.conf` file should be used during
 translation.
 
-    mgmt run --puppet /opt/my-manifest.pp --puppet-conf /etc/mgmt/puppet.conf
+```
+mgmt run --puppet /opt/my-manifest.pp --puppet-conf /etc/mgmt/puppet.conf
+```
 
 Within this file, you can just specify any needed options in the
 `[main]` section:
 
-    [main]
-    server=mgmt-master.example.net
-    vardir=/var/lib/mgmt/puppet
+```
+[main]
+server=mgmt-master.example.net
+vardir=/var/lib/mgmt/puppet
+```
 
-##Caveats
+## Caveats
 
 Please see the [README](https://github.com/ffrank/puppet-mgmtgraph/blob/master/README.md)
 of the translator module for the current state of supported and unsupported

docs/quick-start-guide.md

@@ -0,0 +1,93 @@
+# Quick start guide
+
+## Introduction
+This guide is intended for developers. Once `mgmt` is minimally viable, we'll
+publish a quick start guide for users too. In the meantime, please contribute!
+If you're brand new to `mgmt`, it's probably a good idea to start by reading the
+[introductory article](https://ttboj.wordpress.com/2016/01/18/next-generation-configuration-mgmt/)
+or to watch an [introductory video](https://github.com/purpleidea/mgmt/#on-the-web).
+Once you're familiar with the general idea, please start hacking...
+
+## Vagrant
+If you would like to avoid doing the following steps manually, we have prepared
+a [Vagrant](https://www.vagrantup.com/) environment for your convenience. From
+the project directory, run a `vagrant up`, and then a `vagrant status`. From
+there, you can `vagrant ssh` into the `mgmt` machine. The MOTD will explain the
+rest.
+
+## Dependencies
+Software projects have a few different kinds of dependencies. There are _build_
+dependencies, _runtime_ dependencies, and additionally, a few extra dependencies
+required for running the _test_ suite.
+
+### Build
+* `golang` 1.6 or higher (required, available in most distros)
+* golang libraries (required, available with `go get ./...`) a partial list includes:
+```
+github.com/coreos/etcd/client
+gopkg.in/yaml.v2
+gopkg.in/fsnotify.v1
+github.com/urfave/cli
+github.com/coreos/go-systemd/dbus
+github.com/coreos/go-systemd/util
+github.com/libvirt/libvirt-go
+```
+* `stringer` (optional), available as a package on some platforms, otherwise via `go get`
+```
+golang.org/x/tools/cmd/stringer
+```
+* `pandoc` (optional), for building a pdf of the documentation
+
+### Runtime
+A relatively modern GNU/Linux system should be able to run `mgmt` without any
+problems. Since `mgmt` runs as a single statically compiled binary, all of the
+library dependencies are included. It is expected, that certain advanced
+resources require host specific facilities to work. These requirements are
+listed below:
+
+| Resource | Dependency        | Version |
+|----------|-------------------|---------|
+| file     | inotify           | ?       |
+| hostname | systemd-hostnamed | ?       |
+| nspawn   | systemd-nspawn    | ?       |
+| pkg      | packagekitd       | ?       |
+| svc      | systemd           | ?       |
+| virt     | libvirtd          | ?       |
+
+For building a visual representation of the graph, `graphviz` is required.
+
+### Testing
+* golint `github.com/golang/lint/golint`
+
+## Quick start
+* Make sure you have golang version 1.6 or greater installed.
+* If you do not have a GOPATH yet, create one and export it:
+```
+mkdir $HOME/gopath
+export GOPATH=$HOME/gopath
+```
+* You might also want to add the GOPATH to your `~/.bashrc` or `~/.profile`.
+* For more information you can read the [GOPATH documentation](https://golang.org/cmd/go/#hdr-GOPATH_environment_variable).
+* Next download the mgmt code base, and switch to that directory:
+```
+mkdir -p $GOPATH/src/github.com/purpleidea/
+cd $GOPATH/src/github.com/purpleidea/
+git clone --recursive https://github.com/purpleidea/mgmt/
+cd $GOPATH/src/github.com/purpleidea/mgmt
+```
+* Run `make deps` to install system and golang dependencies. Take a look at `misc/make-deps.sh` for details.
+* Run `make build` to get a freshly built `mgmt` binary.
+* Run `time ./mgmt run --yaml examples/graph0.yaml --converged-timeout=5 --tmp-prefix` to try out a very simple example!
+* To run continuously in the default mode of operation, omit the `--converged-timeout` option.
+* Have fun hacking on our future technology!
+
+## Examples
+Please look in the [examples/](../examples/) folder for some examples!
+
+## Installation
+Installation of `mgmt` from distribution packages currently needs improvement.
+At the moment we have:
+* [COPR](https://copr.fedoraproject.org/coprs/purpleidea/mgmt/)
+* [Arch](https://aur.archlinux.org/packages/mgmt/)
+
+Please contribute more! We'd especially like to see a Debian package!

docs/resource-guide.md

@@ -0,0 +1,575 @@
+# Resource guide
+
+## Overview
+
+The `mgmt` tool has built-in resource primitives which make up the building
+blocks of any configuration. Each instance of a resource is mapped to a single
+vertex in the resource [graph](https://en.wikipedia.org/wiki/Directed_acyclic_graph).
+This guide is meant to instruct developers on how to write a brand new resource.
+Since `mgmt` and the core resources are written in golang, some prior golang
+knowledge is assumed.
+
+## Theory
+
+Resources in `mgmt` are similar to resources in other systems in that they are
+[idempotent](https://en.wikipedia.org/wiki/Idempotence). Our resources are
+uniquely different in that they can detect when their state has changed, and as
+a result can run to revert or repair this change instantly. For some background
+on this design, please read the
+[original article](https://ttboj.wordpress.com/2016/01/18/next-generation-configuration-mgmt/)
+on the subject.
+
+## Resource API
+
+To implement a resource in `mgmt` it must satisfy the
+[`Res`](https://github.com/purpleidea/mgmt/blob/master/resources/resources.go)
+interface. What follows are each of the method signatures and a description of
+each.
+
+### Default
+```golang
+Default() Res
+```
+
+This returns a populated resource struct as a `Res`. It shouldn't populate any
+values which already have the correct default as the golang zero value. In
+general it is preferable if the zero values make for the correct defaults.
+
+#### Example
+```golang
+// Default returns some sensible defaults for this resource.
+func (obj *FooRes) Default() Res {
+	return &FooRes{
+		Answer: 42, // sometimes, defaults shouldn't be the zero value
+	}
+}
+```
+
+### Validate
+```golang
+Validate() error
+```
+
+This method is used to validate if the populated resource struct is a valid
+representation of the resource kind. If it does not conform to the resource
+specifications, it should generate an error. If you notice that this method is
+quite large, it might be an indication that you should reconsider the parameter
+list and interface to this resource. This method is called _before_ `Init`.
+
+#### Example
+```golang
+// Validate reports any problems with the struct definition.
+func (obj *FooRes) Validate() error {
+	if obj.Answer != 42 { // validate whatever you want
+		return fmt.Errorf("expected an answer of 42")
+	}
+	return obj.BaseRes.Validate() // remember to call the base method!
+}
+```
+
+### Init
+```golang
+Init() error
+```
+
+This is called to initialize the resource. If something goes wrong, it should
+return an error. It should set the resource `kind`, do any resource specific
+work, and finish by calling the `Init` method of the base resource.
+
+#### Example
+```golang
+// Init initializes the Foo resource.
+func (obj *FooRes) Init() error {
+	obj.BaseRes.kind = "foo" // must lower case resource kind
+	// run the resource specific initialization, and error if anything fails
+	if some_error {
+		return err // something went wrong!
+	}
+	return obj.BaseRes.Init() // call the base resource init
+}
+```
+
+This method is always called after `Validate` has run successfully, with the
+exception that we can't prevent a malicious or buggy `libmgmt` user to not run
+this. In other words, you should expect `Validate` to have run first, but you
+shouldn't allow `Init` to dangerously `rm -rf /$the_world` if your code only
+checks `$the_world` in `Validate`. Remember to always program safely!
+
+### Close
+```golang
+Close() error
+```
+
+This is called to cleanup after the resource. It is usually not necessary, but
+can be useful if you'd like to properly close a persistent connection that you
+opened in the `Init` method and were using throughout the resource.
+
+#### Example
+```golang
+// Close runs some cleanup code for this resource.
+func (obj *FooRes) Close() error {
+	err := obj.conn.Close() // close some internal connection
+
+	// call base close, b/c we're overriding
+	if e := obj.BaseRes.Close(); err == nil {
+		err = e
+	} else if e != nil {
+		err = multierr.Append(err, e) // list of errors
+	}
+	return err
+}
+```
+
+You should probably check the return errors of your internal methods, and pass
+on an error if something went wrong. Remember to always call the base `Close`
+method! If you plan to return early if you hit an internal error, then at least
+call it with a defer!
+
+### CheckApply
+```golang
+CheckApply(apply bool) (checkOK bool, err error)
+```
+
+`CheckApply` is where the real _work_ is done. Under normal circumstances, this
+function should check if the state of this resource is correct, and if so, it
+should return: `(true, nil)`. If the `apply` variable is set to `true`, then
+this means that we should then proceed to run the changes required to bring the
+resource into the correct state. If the `apply` variable is set to `false`, then
+the resource is operating in _noop_ mode and _no operations_ should be executed!
+
+After having executed the necessary operations to bring the resource back into
+the desired state, or after having detected that the state was incorrect, but
+that changes can't be made because `apply` is `false`, you should then return
+`(false, nil)`.
+
+You must cause the resource to converge during a single execution of this
+function. If you cannot, then you must return an error! The exception to this
+rule is that if an external force changes the state of the resource while it is
+being remedied, it is possible to return from this function even though the
+resource isn't now converged. This is not a bug, as the resources `Watch`
+facility will detect the change, ultimately resulting in a subsequent call to
+`CheckApply`.
+
+#### Example
+```golang
+// CheckApply does the idempotent work of checking and applying resource state.
+func (obj *FooRes) CheckApply(apply bool) (bool, error) {
+	// check the state
+	if state_is_okay { return true, nil } // done early! :)
+	// state was bad
+	if !apply { return false, nil } // don't apply; !stateok, nil
+	// do the apply!
+	return false, nil // after success applying
+	if any_error { return false, err } // anytime there's an err!
+}
+```
+
+The `CheckApply` function is called by the `mgmt` engine when it believes a call
+is necessary. Under certain conditions when a `Watch` call does not invalidate
+the state of the resource, and no refresh call was sent, its execution might be
+skipped. This is an engine optimization, and not a bug. It is mentioned here in
+the documentation in case you are confused as to why a debug message you've
+added to the code isn't always printed.
+
+#### Refresh notifications
+Some resources may choose to support receiving refresh notifications. In general
+these should be avoided if possible, but nevertheless, they do make sense in
+certain situations. Resources that support these need to verify if one was sent
+during the CheckApply phase of execution. This is accomplished by calling the
+`Refresh() bool` method of the resource, and inspecting the return value. This
+is only necessary if you plan to perform a refresh action. Refresh actions
+should still respect the `apply` variable, and no system changes should be made
+if it is `false`. Refresh notifications are generated by any resource when an
+action is applied by that resource and are transmitted through graph edges which
+have enabled their propagation. Resources that currently perform some refresh
+action include `svc`, `timer`, and `password`.
+
+#### Paired execution
+For many resources it is not uncommon to see `CheckApply` run twice in rapid
+succession. This is usually not a pathological occurrence, but rather a healthy
+pattern which is a consequence of the event system. When the state of the
+resource is incorrect, `CheckApply` will run to remedy the state. In response to
+having just changed the state, it is usually the case that this repair will
+trigger the `Watch` code! In response, a second `CheckApply` is triggered, which
+will likely find the state to now be correct.
+
+#### Summary
+* Anytime an error occurs during `CheckApply`, you should return `(false, err)`.
+* If the state is correct and no changes are needed, return `(true, nil)`.
+* You should only make changes to the system if `apply` is set to `true`.
+* After checking the state and possibly applying the fix, return `(false, nil)`.
+* Returning `(true, err)` is a programming error and will cause a `Fatal`.
+
+### Watch
+```golang
+Watch(chan *Event) error
+```
+
+`Watch` is a main loop that runs and sends messages when it detects that the
+state of the resource might have changed. To send a message you should write to
+the input event channel using the `Event` helper method. The Watch function
+should run continuously until a shutdown message is received. If at any time
+something goes wrong, you should return an error, and the `mgmt` engine will
+handle possibly restarting the main loop based on the `retry` meta parameters.
+
+It is better to send an event notification which turns out to be spurious, than
+to miss a possible event. Resources which can miss events are incorrect and need
+to be re-engineered so that this isn't the case. If you have an idea for a
+resource which would fit this criteria, but you can't find a solution, please
+contact the `mgmt` maintainers so that this problem can be investigated and a
+possible system level engineering fix can be found.
+
+You may have trouble deciding how much resource state checking should happen in
+the `Watch` loop versus deferring it all to the `CheckApply` method. You may
+want to put some simple fast path checking in `Watch` to avoid generating
+obviously spurious events, but in general it's best to keep the `Watch` method
+as simple as possible. Contact the `mgmt` maintainers if you're not sure.
+
+If the resource is activated in `polling` mode, the `Watch` method will not get
+executed. As a result, the resource must still work even if the main loop is not
+running.
+
+#### Select
+The lifetime of most resources `Watch` method should be spent in an infinite
+loop that is bounded by a `select` call. The `select` call is the point where
+our method hands back control to the engine (and the kernel) so that we can
+sleep until something of interest wakes us up. In this loop we must process
+events from the engine via the `<-obj.Events()` call, and receive events for our
+resource itself!
+
+#### Events
+If we receive an internal event from the `<-obj.Events()` method, we can read it
+with the ReadEvent helper function. This function tells us if we should shutdown
+our resource, and if we should generate an event. When we want to send an event,
+we use the `Event` helper function. It is also important to mark the resource
+state as `dirty` if we believe it might have changed. We do this with the
+`StateOK(false)` function.
+
+#### Startup
+Once the `Watch` function has finished starting up successfully, it is important
+to generate one event to notify the `mgmt` engine that we're now listening
+successfully, so that it can run an initial `CheckApply` to ensure we're safely
+tracking a healthy state and that we didn't miss anything when `Watch` was down
+or from before `mgmt` was running. It does this by calling the `Running` method.
+
+#### Converged
+The engine might be asked to shutdown when the entire state of the system has
+not seen any changes for some duration of time. The engine can determine this
+automatically, but each resource can block this if it is absolutely necessary.
+To do this, the `Watch` method should get the `ConvergedUID` handle that has
+been prepared for it by the engine. This is done by calling the `ConvergerUID`
+method on the resource object. The result can be used to set the converged
+status with `SetConverged`, and to notify when the particular timeout has been
+reached by waiting on `ConvergedTimer`.
+
+Instead of interacting with the `ConvergedUID` with these two methods, we can
+instead use the `StartTimer` and `ResetTimer` methods which accomplish the same
+thing, but provide a `select`-free interface for different coding situations.
+
+This particular facility is most likely not required for most resources. It may
+prove to be useful if a resource wants to start off a long operation, but avoid
+sending out erroneous `Event` messages to keep things alive until it finishes.
+
+#### Example
+```golang
+// Watch is the listener and main loop for this resource.
+func (obj *FooRes) Watch() error {
+	// setup the Foo resource
+	var err error
+	if err, obj.foo = OpenFoo(); err != nil {
+		return err // we couldn't startup
+	}
+	defer obj.whatever.CloseFoo() // shutdown our
+
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
+	var send = false // send event?
+	var exit *error
+	for {
+		select {
+		case event := <-obj.Events():
+			// we avoid sending events on unpause
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
+			}
+
+		// the actual events!
+		case event := <-obj.foo.Events:
+			if is_an_event {
+				send = true // used below
+				obj.StateOK(false) // dirty
+			}
+
+		// event errors
+		case err := <-obj.foo.Errors:
+			return err // will cause a retry or permanent failure
+		}
+
+		// do all our event sending all together to avoid duplicate msgs
+		if send {
+			send = false
+			obj.Event() // send the event!
+		}
+	}
+}
+```
+
+#### Summary
+* Remember to call the appropriate `converger` methods throughout the resource.
+* Remember to call `Startup` when the `Watch` is running successfully.
+* Remember to process internal events and shutdown promptly if asked to.
+* Ensure the design of your resource is well thought out.
+* Have a look at the existing resources for a rough idea of how this all works.
+
+### Compare
+```golang
+Compare(Res) bool
+```
+
+Each resource must have a `Compare` method. This takes as input another resource
+and must return whether they are identical or not. This is used for identifying
+if an existing resource can be used in place of a new one with a similar set of
+parameters. In particular, when switching from one graph to a new (possibly
+identical) graph, this avoids recomputing the state for resources which don't
+change or that are sufficiently similar that they don't need to be swapped out.
+
+In general if all the resource properties are identical, then they usually don't
+need to be changed. On occasion, not all of them need to be compared, in
+particular if they store some generated state, or if they aren't significant in
+some way.
+
+#### Example
+```golang
+// Compare two resources and return if they are equivalent.
+func (obj *FooRes) Compare(res Res) bool {
+	switch res.(type) {
+	case *FooRes: // only compare to other resources of the Foo kind!
+		res := res.(*FileRes)
+		if !obj.BaseRes.Compare(res) { // call base Compare
+			return false
+		}
+		if obj.Name != res.Name {
+			return false
+		}
+		if obj.whatever != res.whatever {
+			return false
+		}
+		if obj.Flag != res.Flag {
+			return false
+		}
+	default:
+		return false // different kind of resource
+	}
+	return true // they must match!
+}
+```
+
+### UIDs
+```golang
+UIDs() []ResUID
+```
+
+The `UIDs` method returns a list of `ResUID` interfaces that represent the
+particular resource uniquely. This is used with the AutoEdges API to determine
+if another resource can match a dependency to this one.
+
+### AutoEdges
+```golang
+AutoEdges() AutoEdge
+```
+
+This returns a struct that implements the `AutoEdge` interface. This struct
+is used to match other resources that might be relevant dependencies for this
+resource.
+
+### CollectPattern
+```golang
+CollectPattern() string
+```
+
+This is currently a stub and will be updated once the DSL is further along.
+
+### UnmarshalYAML
+```golang
+UnmarshalYAML(unmarshal func(interface{}) error) error // optional
+```
+
+This is optional, but recommended for any resource that will have a YAML
+accessible struct, and an entry in the `GraphConfig` struct. It is not required
+because to do so would mean that third-party or custom resources (such as those
+someone writes to use with `libmgmt`) would have to implement this needlessly.
+
+The signature intentionally matches what is required to satisfy the `go-yaml`
+[Unmarshaler](https://godoc.org/gopkg.in/yaml.v2#Unmarshaler) interface.
+
+#### Example
+```golang
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *FooRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes FooRes // indirection to avoid infinite recursion
+
+	def := obj.Default()     // get the default
+	res, ok := def.(*FooRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to FooRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = FooRes(raw) // restore from indirection with type conversion!
+	return nil
+}
+```
+
+## Further considerations
+There is some additional information that any resource writer will need to know.
+Each issue is listed separately below!
+
+### Resource struct
+Each resource will implement methods as pointer receivers on a resource struct.
+The resource struct must include an anonymous reference to the `BaseRes` struct.
+The naming convention for resources is that they end with a `Res` suffix. If
+you'd like your resource to be accessible by the `YAML` graph API (GAPI), then
+you'll need to include the appropriate YAML fields as shown below.
+
+#### Example
+```golang
+type FooRes struct {
+	BaseRes `yaml:",inline"` // base properties
+
+	Whatever string `yaml:"whatever"` // you pick!
+	Bar int // no yaml, used as public output value for send/recv
+	Baz bool `yaml:"baz"` // something else
+
+	something string // some private field
+}
+```
+
+### YAML
+In addition to labelling your resource struct with YAML fields, you must also
+add an entry to the internal `GraphConfig` struct. It is a fairly straight
+forward one line patch.
+
+```golang
+type GraphConfig struct {
+// [snip...]
+	Resources struct {
+		Noop []*resources.NoopRes `yaml:"noop"`
+		File []*resources.FileRes `yaml:"file"`
+		// [snip...]
+		Foo []*resources.FooRes `yaml:"foo"` // tada :)
+	}
+}
+```
+
+It's also recommended that you add the [UnmarshalYAML](#unmarshalyaml) method to
+your resources so that unspecified values are given sane defaults.
+
+### Gob registration
+All resources must be registered with the `golang` _gob_ module so that they can
+be encoded and decoded. Make sure to include the following code snippet for this
+to work.
+
+```golang
+import "encoding/gob"
+func init() { // special golang method that runs once
+	gob.Register(&FooRes{}) // substitude your resource here
+}
+```
+
+## Automatic edges
+Automatic edges in `mgmt` are well described in [this article](https://ttboj.wordpress.com/2016/03/14/automatic-edges-in-mgmt/).
+The best example of this technique can be seen in the `svc` resource.
+Unfortunately no further documentation about this subject has been written. To
+expand this section, please send a patch! Please contact us if you'd like to
+work on a resource that uses this feature, or to add it to an existing one!
+
+## Automatic grouping
+Automatic grouping in `mgmt` is well described in [this article](https://ttboj.wordpress.com/2016/03/30/automatic-grouping-in-mgmt/).
+The best example of this technique can be seen in the `pkg` resource.
+Unfortunately no further documentation about this subject has been written. To
+expand this section, please send a patch! Please contact us if you'd like to
+work on a resource that uses this feature, or to add it to an existing one!
+
+
+## Send/Recv
+In `mgmt` there is a novel concept called _Send/Recv_. For some background,
+please [read the introductory article](https://ttboj.wordpress.com/2016/12/07/sendrecv-in-mgmt/).
+When using this feature, the engine will automatically send the user specified
+value to the intended destination without requiring any resource specific code.
+Any time that one of the destination values is changed, the engine automatically
+marks the resource state as `dirty`. To detect if a particular value was
+received, and if it changed (during this invocation of CheckApply) from the
+previous value, you can query the Recv parameter. It will contain a `map` of all
+the keys which can be received on, and the value has a `Changed` property which
+will indicate whether the value was updated on this particular `CheckApply`
+invocation. The type of the sending key must match that of the receiving one.
+This can _only_ be done inside of the `CheckApply` function!
+
+```golang
+// inside CheckApply, probably near the top
+if val, exists := obj.Recv["SomeKey"]; exists {
+	log.Printf("SomeKey was sent to us from: %s[%s].%s", val.Res.Kind(), val.Res.GetName(), val.Key)
+	if val.Changed {
+		log.Printf("SomeKey was just updated!")
+		// you may want to invalidate some local cache
+	}
+}
+```
+
+Astute readers will note that there isn't anything that prevents a user from
+sending an identically typed value to some arbitrary (public) key that the
+resource author hadn't considered! While this is true, resources should probably
+work within this problem space anyways. The rule of thumb is that any public
+parameter which is normally used in a resource can be used safely.
+
+One subtle scenario is that if a resource creates a local cache or stores a
+computation that depends on the value of a public parameter and will require
+invalidation should that public parameter change, then you must detect that
+scenario and invalidate the cache when it occurs. This *must* be processed
+before there is a possibility of failure in CheckApply, because if we fail (and
+possibly run again) the subsequent send->recv transfer might not have a new
+value to copy, and therefore we won't see this notification of change.
+Therefore, it is important to process these promptly, if they must not be lost,
+such as for cache invalidation.
+
+Remember, `Send/Recv` only changes your resource code if you cache state.
+
+## Composite resources
+Composite resources are resources which embed one or more existing resources.
+This is useful to prevent code duplication in higher level resource scenarios.
+The best example of this technique can be seen in the `nspawn` resource which
+can be seen to partially embed a `svc` resource, but without its `Watch`.
+Unfortunately no further documentation about this subject has been written. To
+expand this section, please send a patch! Please contact us if you'd like to
+work on a resource that uses this feature, or to add it to an existing one!
+
+## Frequently asked questions
+(Send your questions as a patch to this FAQ! I'll review it, merge it, and
+respond by commit with the answer.)
+
+### Can I write resources in a different language?
+Currently `golang` is the only supported language for built-in resources. We
+might consider allowing external resources to be imported in the future. This
+will likely require a language that can expose a C-like API, such as `python` or
+`ruby`. Custom `golang` resources are already possible when using mgmt as a lib.
+Higher level resource collections will be possible once the `mgmt` DSL is ready.
+
+### What new resource primitives need writing?
+There are still many ideas for new resources that haven't been written yet. If
+you'd like to contribute one, please contact us and tell us about your idea!
+
+### Where can I find more information about mgmt?
+Additional blog posts, videos and other material [is available!](https://github.com/purpleidea/mgmt/#on-the-web).
+
+## Suggestions
+If you have any ideas for API changes or other improvements to resource writing,
+please let us know! We're still pre 1.0 and pre 0.1 and happy to break API in
+order to get it right!

etcd/etcd.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -37,11 +37,12 @@
 //
 // Smoke testing:
 // mkdir /tmp/mgmt{A..E}
-// ./mgmt run --yaml examples/etcd1a.yaml --hostname h1 --tmp-prefix
-// ./mgmt run --yaml examples/etcd1b.yaml --hostname h2 --tmp-prefix --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2381 --server-urls http://127.0.0.1:2382
-// ./mgmt run --yaml examples/etcd1c.yaml --hostname h3 --tmp-prefix --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2383 --server-urls http://127.0.0.1:2384
+// ./mgmt run --yaml examples/etcd1a.yaml --hostname h1 --tmp-prefix --no-pgp
+// ./mgmt run --yaml examples/etcd1b.yaml --hostname h2 --tmp-prefix --no-pgp --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2381 --server-urls http://127.0.0.1:2382
+// ./mgmt run --yaml examples/etcd1c.yaml --hostname h3 --tmp-prefix --no-pgp --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2383 --server-urls http://127.0.0.1:2384
 // ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2379 put /_mgmt/idealClusterSize 3
-// ./mgmt run --yaml examples/etcd1d.yaml --hostname h4 --tmp-prefix --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2385 --server-urls http://127.0.0.1:2386
+// ./mgmt run --yaml examples/etcd1d.yaml --hostname h4 --tmp-prefix --no-pgp --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2385 --server-urls http://127.0.0.1:2386
+// ./mgmt run --yaml examples/etcd1e.yaml --hostname h5 --tmp-prefix --no-pgp --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2387 --server-urls http://127.0.0.1:2388
 // ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2379 member list
 // ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2381 put /_mgmt/idealClusterSize 5
 // ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2381 member list
@@ -92,7 +93,7 @@ const (
 )
 
 var (
-	errApplyDeltaEventsInconsistent = errors.New("Etcd: ApplyDeltaEvents: Inconsistent key!")
+	errApplyDeltaEventsInconsistent = errors.New("inconsistent key in ApplyDeltaEvents")
 )
 
 // AW is a struct for the AddWatcher queue
@@ -316,7 +317,7 @@ func (obj *EmbdEtcd) Connect(reconnect bool) error {
 			if emax > maxClientConnectRetries {
 				log.Printf("Etcd: The dataDir (%s) might be inconsistent or corrupt.", obj.dataDir)
 				log.Printf("Etcd: Please see: %s", "https://github.com/purpleidea/mgmt/blob/master/DOCUMENTATION.md#what-does-the-error-message-about-an-inconsistent-datadir-mean")
-				obj.cError = fmt.Errorf("Can't find an available endpoint.")
+				obj.cError = fmt.Errorf("can't find an available endpoint")
 				return obj.cError
 			}
 			err = &CtxDelayErr{time.Duration(emax) * time.Second, "No endpoints available yet!"} // retry with backoff...
@@ -367,7 +368,7 @@ func (obj *EmbdEtcd) Startup() error {
 	// if we have no endpoints, it means we are bootstrapping...
 	if !bootstrapping {
 		log.Println("Etcd: Startup: Getting initial values...")
-		if nominated, err := EtcdNominated(obj); err == nil {
+		if nominated, err := Nominated(obj); err == nil {
 			obj.nominated = nominated // store a local copy
 		} else {
 			log.Printf("Etcd: Startup: Nominate lookup error.")
@@ -376,7 +377,7 @@ func (obj *EmbdEtcd) Startup() error {
 		}
 
 		// get initial ideal cluster size
-		if idealClusterSize, err := EtcdGetClusterSize(obj); err == nil {
+		if idealClusterSize, err := GetClusterSize(obj); err == nil {
 			obj.idealClusterSize = idealClusterSize
 			log.Printf("Etcd: Startup: Ideal cluster size is: %d", idealClusterSize)
 		} else {
@@ -395,7 +396,7 @@ func (obj *EmbdEtcd) Startup() error {
 	if !obj.noServer && bootstrapping {
 		log.Printf("Etcd: Bootstrapping...")
 		// give an initial value to the obj.nominate map we keep in sync
-		// this emulates EtcdNominate(obj, obj.hostname, obj.serverURLs)
+		// this emulates Nominate(obj, obj.hostname, obj.serverURLs)
 		obj.nominated[obj.hostname] = obj.serverURLs // initial value
 		// NOTE: when we are stuck waiting for the server to start up,
 		// it is probably happening on this call right here...
@@ -406,11 +407,11 @@ func (obj *EmbdEtcd) Startup() error {
 	if !obj.noServer && len(obj.serverURLs) > 0 {
 		// we run this in a go routine because it blocks waiting for server
 		log.Printf("Etcd: Startup: Volunteering...")
-		go EtcdVolunteer(obj, obj.serverURLs)
+		go Volunteer(obj, obj.serverURLs)
 	}
 
 	if bootstrapping {
-		if err := EtcdSetClusterSize(obj, obj.idealClusterSize); err != nil {
+		if err := SetClusterSize(obj, obj.idealClusterSize); err != nil {
 			log.Printf("Etcd: Startup: Ideal cluster size storage error.")
 			obj.Destroy()
 			return fmt.Errorf("Etcd: Startup: Error: %v", err)
@@ -431,7 +432,7 @@ func (obj *EmbdEtcd) Destroy() error {
 
 	// this should also trigger an unnominate, which should cause a shutdown
 	log.Printf("Etcd: Destroy: Unvolunteering...")
-	if err := EtcdVolunteer(obj, nil); err != nil { // unvolunteer so we can shutdown...
+	if err := Volunteer(obj, nil); err != nil { // unvolunteer so we can shutdown...
 		log.Printf("Etcd: Destroy: Error: %v", err) // we have a problem
 	}
 
@@ -578,7 +579,7 @@ func (obj *EmbdEtcd) CtxError(ctx context.Context, err error) (context.Context,
 		log.Fatal("Etcd: CtxError: Error: Unexpected lack of error!")
 	}
 	if obj.exiting {
-		obj.ctxErr = fmt.Errorf("Etcd: CtxError: Exit in progress!")
+		obj.ctxErr = fmt.Errorf("exit in progress")
 		return ctx, obj.ctxErr
 	}
 
@@ -599,14 +600,14 @@ func (obj *EmbdEtcd) CtxError(ctx context.Context, err error) (context.Context,
 	if retriesErr, ok := err.(*CtxRetriesErr); ok { // custom retry error
 		log.Printf("Etcd: CtxError: Reason: %s", retriesErr.Error())
 		if retriesErr.Retries == 0 {
-			obj.ctxErr = fmt.Errorf("Etcd: CtxError: CtxRetriesErr: No more retries!")
+			obj.ctxErr = fmt.Errorf("no more retries due to CtxRetriesErr")
 			return ctx, obj.ctxErr
 		}
 		return ctx, nil
 	}
 
 	if permanentErr, ok := err.(*CtxPermanentErr); ok { // custom permanent error
-		obj.ctxErr = fmt.Errorf("Etcd: CtxError: Reason: %s", permanentErr.Error())
+		obj.ctxErr = fmt.Errorf("error due to CtxPermanentErr: %s", permanentErr.Error())
 		return ctx, obj.ctxErr // quit
 	}
 
@@ -626,7 +627,7 @@ func (obj *EmbdEtcd) CtxError(ctx context.Context, err error) (context.Context,
 	}
 
 	if err == grpc.ErrClientConnTimeout { // sometimes caused by "too many colons" misconfiguration
-		return ctx, fmt.Errorf("Etcd: Error: Misconfiguration: %v", err) // permanent failure?
+		return ctx, fmt.Errorf("misconfiguration: %v", err) // permanent failure?
 	}
 
 	// this can happen if my client connection shuts down, but without any
@@ -670,7 +671,7 @@ func (obj *EmbdEtcd) CtxError(ctx context.Context, err error) (context.Context,
 		log.Printf("Etcd: CtxError: Reconnecting...")
 		if err := obj.Connect(true); err != nil {
 			defer obj.rLock.Unlock()
-			obj.ctxErr = fmt.Errorf("Etcd: Permanent connect error: %v", err)
+			obj.ctxErr = fmt.Errorf("permanent connect error: %v", err)
 			return ctx, obj.ctxErr
 		}
 		if obj.flags.Debug {
@@ -694,7 +695,7 @@ func (obj *EmbdEtcd) CtxError(ctx context.Context, err error) (context.Context,
 	// if you hit this code path here, please report the unmatched error!
 	log.Printf("Etcd: CtxError: Unknown error(%T): %+v", err, err)
 	time.Sleep(1 * time.Second)
-	obj.ctxErr = fmt.Errorf("Etcd: CtxError: Unknown error!")
+	obj.ctxErr = fmt.Errorf("unknown CtxError")
 	return ctx, obj.ctxErr
 }
 
@@ -1188,7 +1189,7 @@ func (obj *EmbdEtcd) volunteerCallback(re *RE) error {
 		return nil // if we're not a server, we're not a leader, return
 	}
 
-	membersMap, err := EtcdMembers(obj) // map[uint64]string
+	membersMap, err := Members(obj) // map[uint64]string
 	if err != nil {
 		return fmt.Errorf("Etcd: Members: Error: %+v", err)
 	}
@@ -1211,7 +1212,7 @@ func (obj *EmbdEtcd) volunteerCallback(re *RE) error {
 		}
 	}
 
-	leader, err := EtcdLeader(obj) // XXX: race!
+	leader, err := Leader(obj) // XXX: race!
 	if err != nil {
 		log.Printf("Etcd: Leader: Error: %+v", err)
 		return fmt.Errorf("Etcd: Leader: Error: %+v", err)
@@ -1224,7 +1225,7 @@ func (obj *EmbdEtcd) volunteerCallback(re *RE) error {
 	// i am the leader!
 
 	// get the list of available volunteers
-	volunteersMap, err := EtcdVolunteers(obj)
+	volunteersMap, err := Volunteers(obj)
 	if err != nil {
 		log.Printf("Etcd: Volunteers: Error: %+v", err)
 		return fmt.Errorf("Etcd: Volunteers: Error: %+v", err)
@@ -1244,7 +1245,7 @@ func (obj *EmbdEtcd) volunteerCallback(re *RE) error {
 	// if we're the only member left, just shutdown...
 	if len(members) == 1 && members[0] == obj.hostname && len(quitters) == 1 && quitters[0] == obj.hostname {
 		log.Printf("Etcd: Quitters: Shutting down self...")
-		if err := EtcdNominate(obj, obj.hostname, nil); err != nil { // unnominate myself
+		if err := Nominate(obj, obj.hostname, nil); err != nil { // unnominate myself
 			return &CtxDelayErr{1 * time.Second, fmt.Sprintf("error shutting down self: %v", err)}
 		}
 		return nil
@@ -1266,7 +1267,7 @@ func (obj *EmbdEtcd) volunteerCallback(re *RE) error {
 		// NOTE: storing peerURLs when they're already in volunteers/ is
 		// redundant, but it seems to be necessary for a sane algorithm.
 		// nominate before we call the API so that members see it first!
-		EtcdNominate(obj, chosen, peerURLs)
+		Nominate(obj, chosen, peerURLs)
 		// XXX: add a ttl here, because once we nominate someone, we
 		// need to give them up to N seconds to start up after we run
 		// the MemberAdd API because if they don't, in some situations
@@ -1279,7 +1280,7 @@ func (obj *EmbdEtcd) volunteerCallback(re *RE) error {
 		// member at a time!
 
 		log.Printf("Etcd: Member Add: %v", peerURLs)
-		mresp, err := EtcdMemberAdd(obj, peerURLs)
+		mresp, err := MemberAdd(obj, peerURLs)
 		if err != nil {
 			// on error this function will run again, which is good
 			// because we need to make sure to run the below parts!
@@ -1310,7 +1311,7 @@ func (obj *EmbdEtcd) volunteerCallback(re *RE) error {
 			// programming error
 			log.Fatalf("Etcd: Member Remove: Error: %v(%v) not in members list!", quitter, mID)
 		}
-		EtcdNominate(obj, quitter, nil) // unnominate
+		Nominate(obj, quitter, nil) // unnominate
 		// once we issue the above unnominate, that peer will
 		// shutdown, and this might cause us to loose quorum,
 		// therefore, let that member remove itself, and then
@@ -1325,7 +1326,7 @@ func (obj *EmbdEtcd) volunteerCallback(re *RE) error {
 		log.Printf("Etcd: Waiting %d seconds for %s to self remove...", selfRemoveTimeout, quitter)
 		time.Sleep(selfRemoveTimeout * time.Second)
 		// in case the removed member doesn't remove itself, do it!
-		removed, err := EtcdMemberRemove(obj, mID)
+		removed, err := MemberRemove(obj, mID)
 		if err != nil {
 			return fmt.Errorf("Etcd: Member Remove: Error: %+v", err)
 		}
@@ -1372,7 +1373,7 @@ func (obj *EmbdEtcd) nominateCallback(re *RE) error {
 		// leader once the current leader calls the MemberAdd API and it
 		// steps down trying to form a two host cluster. Instead, we can
 		// look at the event response data to read the nominated values!
-		//nominated, err = EtcdNominated(obj) // nope, won't always work
+		//nominated, err = Nominated(obj) // nope, won't always work
 		// since we only see what has *changed* in the response data, we
 		// have to keep track of the original state and apply the deltas
 		// this must be idempotent in case it errors and is called again
@@ -1386,7 +1387,7 @@ func (obj *EmbdEtcd) nominateCallback(re *RE) error {
 
 	} else {
 		// TODO: should we just use the above delta method for everything?
-		//nominated, err := EtcdNominated(obj) // just get it
+		//nominated, err := Nominated(obj) // just get it
 		//if err != nil {
 		//	return fmt.Errorf("Etcd: Nominate: Error: %+v", err)
 		//}
@@ -1435,7 +1436,7 @@ func (obj *EmbdEtcd) nominateCallback(re *RE) error {
 
 			// XXX: just put this wherever for now so we don't block
 			// nominate self so "member" list is correct for peers to see
-			EtcdNominate(obj, obj.hostname, obj.serverURLs)
+			Nominate(obj, obj.hostname, obj.serverURLs)
 			// XXX: if this fails, where will we retry this part ?
 		}
 
@@ -1443,7 +1444,7 @@ func (obj *EmbdEtcd) nominateCallback(re *RE) error {
 		if curls := obj.clientURLs; len(curls) > 0 {
 			// XXX: don't advertise local addresses! 127.0.0.1:2381 doesn't really help remote hosts
 			// XXX: but sometimes this is what we want... hmmm how do we decide? filter on callback?
-			EtcdAdvertiseEndpoints(obj, curls)
+			AdvertiseEndpoints(obj, curls)
 			// XXX: if this fails, where will we retry this part ?
 
 			// force this to remove sentinel before we reconnect...
@@ -1454,14 +1455,14 @@ func (obj *EmbdEtcd) nominateCallback(re *RE) error {
 
 	} else if obj.server != nil && !exists {
 		// un advertise client urls
-		EtcdAdvertiseEndpoints(obj, nil)
+		AdvertiseEndpoints(obj, nil)
 
 		// i have been un-nominated, remove self and shutdown server!
 		if len(obj.nominated) != 0 { // don't call if nobody left but me!
 			// this works around: https://github.com/coreos/etcd/issues/5482,
 			// and it probably makes sense to avoid calling if we're the last
 			log.Printf("Etcd: Member Remove: Removing self: %v", obj.memberID)
-			removed, err := EtcdMemberRemove(obj, obj.memberID)
+			removed, err := MemberRemove(obj, obj.memberID)
 			if err != nil {
 				return fmt.Errorf("Etcd: Member Remove: Error: %+v", err)
 			}
@@ -1523,7 +1524,7 @@ func (obj *EmbdEtcd) endpointCallback(re *RE) error {
 	endpoints, err := ApplyDeltaEvents(re, endpoints)
 	if err != nil || exists {
 		// TODO: we could also lookup endpoints from the maintenance api
-		endpoints, err = EtcdEndpoints(obj)
+		endpoints, err = Endpoints(obj)
 		if err != nil {
 			return err
 		}
@@ -1671,16 +1672,16 @@ func (obj *EmbdEtcd) StartServer(newCluster bool, peerURLsMap etcdtypes.URLsMap)
 	case <-obj.server.Server.ReadyNotify(): // we hang here if things are bad
 		log.Printf("Etcd: StartServer: Done starting server!") // it didn't hang!
 	case <-time.After(time.Duration(maxStartServerTimeout) * time.Second):
-		e := fmt.Errorf("Etcd: StartServer: Timeout of %d seconds reached!", maxStartServerTimeout)
-		log.Printf(e.Error())
+		e := fmt.Errorf("timeout of %d seconds reached", maxStartServerTimeout)
+		log.Printf("Etcd: StartServer: %s", e.Error())
 		obj.server.Server.Stop() // trigger a shutdown
 		obj.serverwg.Add(1)      // add for the DestroyServer()
 		obj.DestroyServer()
 		return e
 	// TODO: should we wait for this notification elsewhere?
 	case <-obj.server.Server.StopNotify(): // it's going down now...
-		e := fmt.Errorf("Etcd: StartServer: Received stop notification.")
-		log.Printf(e.Error())
+		e := fmt.Errorf("received stop notification")
+		log.Printf("Etcd: StartServer: %s", e.Error())
 		obj.server.Server.Stop() // trigger a shutdown
 		obj.serverwg.Add(1)      // add for the DestroyServer()
 		obj.DestroyServer()
@@ -1716,11 +1717,11 @@ func (obj *EmbdEtcd) DestroyServer() error {
 // TODO: Could all these Etcd*(obj *EmbdEtcd, ...) functions which deal with the
 // interface between etcd paths and behaviour be grouped into a single struct ?
 
-// EtcdNominate nominates a particular client to be a server (peer)
-func EtcdNominate(obj *EmbdEtcd, hostname string, urls etcdtypes.URLs) error {
+// Nominate nominates a particular client to be a server (peer)
+func Nominate(obj *EmbdEtcd, hostname string, urls etcdtypes.URLs) error {
 	if obj.flags.Trace {
-		log.Printf("Trace: Etcd: EtcdNominate(%v): %v", hostname, urls.String())
-		defer log.Printf("Trace: Etcd: EtcdNominate(%v): Finished!", hostname)
+		log.Printf("Trace: Etcd: Nominate(%v): %v", hostname, urls.String())
+		defer log.Printf("Trace: Etcd: Nominate(%v): Finished!", hostname)
 	}
 	// nominate someone to be a server
 	nominate := fmt.Sprintf("/%s/nominated/%s", NS, hostname)
@@ -1733,18 +1734,18 @@ func EtcdNominate(obj *EmbdEtcd, hostname string, urls etcdtypes.URLs) error {
 	}
 
 	if _, err := obj.Txn(nil, ops, nil); err != nil {
-		return fmt.Errorf("Etcd: Nominate failed!") // exit in progress?
+		return fmt.Errorf("nominate failed") // exit in progress?
 	}
 	return nil
 }
 
-// EtcdNominated returns a urls map of nominated etcd server volunteers
+// Nominated returns a urls map of nominated etcd server volunteers
 // NOTE: I know 'nominees' might be more correct, but is less consistent here
-func EtcdNominated(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
+func Nominated(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 	path := fmt.Sprintf("/%s/nominated/", NS)
 	keyMap, err := obj.Get(path, etcd.WithPrefix()) // map[string]string, bool
 	if err != nil {
-		return nil, fmt.Errorf("Etcd: Nominated isn't available: %v", err)
+		return nil, fmt.Errorf("nominated isn't available: %v", err)
 	}
 	nominated := make(etcdtypes.URLsMap)
 	for key, val := range keyMap { // loop through directory of nominated
@@ -1757,7 +1758,7 @@ func EtcdNominated(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 		}
 		urls, err := etcdtypes.NewURLs(strings.Split(val, ","))
 		if err != nil {
-			return nil, fmt.Errorf("Etcd: Nominated: Data format error!: %v", err)
+			return nil, fmt.Errorf("nominated data format error: %v", err)
 		}
 		nominated[name] = urls // add to map
 		if obj.flags.Debug {
@@ -1767,11 +1768,11 @@ func EtcdNominated(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 	return nominated, nil
 }
 
-// EtcdVolunteer offers yourself up to be a server if needed
-func EtcdVolunteer(obj *EmbdEtcd, urls etcdtypes.URLs) error {
+// Volunteer offers yourself up to be a server if needed
+func Volunteer(obj *EmbdEtcd, urls etcdtypes.URLs) error {
 	if obj.flags.Trace {
-		log.Printf("Trace: Etcd: EtcdVolunteer(%v): %v", obj.hostname, urls.String())
-		defer log.Printf("Trace: Etcd: EtcdVolunteer(%v): Finished!", obj.hostname)
+		log.Printf("Trace: Etcd: Volunteer(%v): %v", obj.hostname, urls.String())
+		defer log.Printf("Trace: Etcd: Volunteer(%v): Finished!", obj.hostname)
 	}
 	// volunteer to be a server
 	volunteer := fmt.Sprintf("/%s/volunteers/%s", NS, obj.hostname)
@@ -1785,21 +1786,21 @@ func EtcdVolunteer(obj *EmbdEtcd, urls etcdtypes.URLs) error {
 	}
 
 	if _, err := obj.Txn(nil, ops, nil); err != nil {
-		return fmt.Errorf("Etcd: Volunteering failed!") // exit in progress?
+		return fmt.Errorf("volunteering failed") // exit in progress?
 	}
 	return nil
 }
 
-// EtcdVolunteers returns a urls map of available etcd server volunteers
-func EtcdVolunteers(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
+// Volunteers returns a urls map of available etcd server volunteers
+func Volunteers(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 	if obj.flags.Trace {
-		log.Printf("Trace: Etcd: EtcdVolunteers()")
-		defer log.Printf("Trace: Etcd: EtcdVolunteers(): Finished!")
+		log.Printf("Trace: Etcd: Volunteers()")
+		defer log.Printf("Trace: Etcd: Volunteers(): Finished!")
 	}
 	path := fmt.Sprintf("/%s/volunteers/", NS)
 	keyMap, err := obj.Get(path, etcd.WithPrefix())
 	if err != nil {
-		return nil, fmt.Errorf("Etcd: Volunteers aren't available: %v", err)
+		return nil, fmt.Errorf("volunteers aren't available: %v", err)
 	}
 	volunteers := make(etcdtypes.URLsMap)
 	for key, val := range keyMap { // loop through directory of volunteers
@@ -1812,7 +1813,7 @@ func EtcdVolunteers(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 		}
 		urls, err := etcdtypes.NewURLs(strings.Split(val, ","))
 		if err != nil {
-			return nil, fmt.Errorf("Etcd: Volunteers: Data format error!: %v", err)
+			return nil, fmt.Errorf("volunteers data format error: %v", err)
 		}
 		volunteers[name] = urls // add to map
 		if obj.flags.Debug {
@@ -1822,11 +1823,11 @@ func EtcdVolunteers(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 	return volunteers, nil
 }
 
-// EtcdAdvertiseEndpoints advertises the list of available client endpoints
-func EtcdAdvertiseEndpoints(obj *EmbdEtcd, urls etcdtypes.URLs) error {
+// AdvertiseEndpoints advertises the list of available client endpoints
+func AdvertiseEndpoints(obj *EmbdEtcd, urls etcdtypes.URLs) error {
 	if obj.flags.Trace {
-		log.Printf("Trace: Etcd: EtcdAdvertiseEndpoints(%v): %v", obj.hostname, urls.String())
-		defer log.Printf("Trace: Etcd: EtcdAdvertiseEndpoints(%v): Finished!", obj.hostname)
+		log.Printf("Trace: Etcd: AdvertiseEndpoints(%v): %v", obj.hostname, urls.String())
+		defer log.Printf("Trace: Etcd: AdvertiseEndpoints(%v): Finished!", obj.hostname)
 	}
 	// advertise endpoints
 	endpoints := fmt.Sprintf("/%s/endpoints/%s", NS, obj.hostname)
@@ -1840,21 +1841,21 @@ func EtcdAdvertiseEndpoints(obj *EmbdEtcd, urls etcdtypes.URLs) error {
 	}
 
 	if _, err := obj.Txn(nil, ops, nil); err != nil {
-		return fmt.Errorf("Etcd: Endpoint advertising failed!") // exit in progress?
+		return fmt.Errorf("endpoint advertising failed") // exit in progress?
 	}
 	return nil
 }
 
-// EtcdEndpoints returns a urls map of available etcd server endpoints
-func EtcdEndpoints(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
+// Endpoints returns a urls map of available etcd server endpoints
+func Endpoints(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 	if obj.flags.Trace {
-		log.Printf("Trace: Etcd: EtcdEndpoints()")
-		defer log.Printf("Trace: Etcd: EtcdEndpoints(): Finished!")
+		log.Printf("Trace: Etcd: Endpoints()")
+		defer log.Printf("Trace: Etcd: Endpoints(): Finished!")
 	}
 	path := fmt.Sprintf("/%s/endpoints/", NS)
 	keyMap, err := obj.Get(path, etcd.WithPrefix())
 	if err != nil {
-		return nil, fmt.Errorf("Etcd: Endpoints aren't available: %v", err)
+		return nil, fmt.Errorf("endpoints aren't available: %v", err)
 	}
 	endpoints := make(etcdtypes.URLsMap)
 	for key, val := range keyMap { // loop through directory of endpoints
@@ -1867,7 +1868,7 @@ func EtcdEndpoints(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 		}
 		urls, err := etcdtypes.NewURLs(strings.Split(val, ","))
 		if err != nil {
-			return nil, fmt.Errorf("Etcd: Endpoints: Data format error!: %v", err)
+			return nil, fmt.Errorf("endpoints data format error: %v", err)
 		}
 		endpoints[name] = urls // add to map
 		if obj.flags.Debug {
@@ -1877,30 +1878,30 @@ func EtcdEndpoints(obj *EmbdEtcd) (etcdtypes.URLsMap, error) {
 	return endpoints, nil
 }
 
-// EtcdSetHostnameConverged sets whether a specific hostname is converged.
-func EtcdSetHostnameConverged(obj *EmbdEtcd, hostname string, isConverged bool) error {
+// SetHostnameConverged sets whether a specific hostname is converged.
+func SetHostnameConverged(obj *EmbdEtcd, hostname string, isConverged bool) error {
 	if obj.flags.Trace {
-		log.Printf("Trace: Etcd: EtcdSetHostnameConverged(%s): %v", hostname, isConverged)
-		defer log.Printf("Trace: Etcd: EtcdSetHostnameConverged(%v): Finished!", hostname)
+		log.Printf("Trace: Etcd: SetHostnameConverged(%s): %v", hostname, isConverged)
+		defer log.Printf("Trace: Etcd: SetHostnameConverged(%v): Finished!", hostname)
 	}
 	converged := fmt.Sprintf("/%s/converged/%s", NS, hostname)
 	op := []etcd.Op{etcd.OpPut(converged, fmt.Sprintf("%t", isConverged))}
 	if _, err := obj.Txn(nil, op, nil); err != nil { // TODO: do we need a skipConv flag here too?
-		return fmt.Errorf("Etcd: Set converged failed!") // exit in progress?
+		return fmt.Errorf("set converged failed") // exit in progress?
 	}
 	return nil
 }
 
-// EtcdHostnameConverged returns a map of every hostname's converged state.
-func EtcdHostnameConverged(obj *EmbdEtcd) (map[string]bool, error) {
+// HostnameConverged returns a map of every hostname's converged state.
+func HostnameConverged(obj *EmbdEtcd) (map[string]bool, error) {
 	if obj.flags.Trace {
-		log.Printf("Trace: Etcd: EtcdHostnameConverged()")
-		defer log.Printf("Trace: Etcd: EtcdHostnameConverged(): Finished!")
+		log.Printf("Trace: Etcd: HostnameConverged()")
+		defer log.Printf("Trace: Etcd: HostnameConverged(): Finished!")
 	}
 	path := fmt.Sprintf("/%s/converged/", NS)
 	keyMap, err := obj.ComplexGet(path, true, etcd.WithPrefix()) // don't un-converge
 	if err != nil {
-		return nil, fmt.Errorf("Etcd: Converged values aren't available: %v", err)
+		return nil, fmt.Errorf("converged values aren't available: %v", err)
 	}
 	converged := make(map[string]bool)
 	for key, val := range keyMap { // loop through directory...
@@ -1913,21 +1914,21 @@ func EtcdHostnameConverged(obj *EmbdEtcd) (map[string]bool, error) {
 		}
 		b, err := strconv.ParseBool(val)
 		if err != nil {
-			return nil, fmt.Errorf("Etcd: Converged: Data format error!: %v", err)
+			return nil, fmt.Errorf("converged data format error: %v", err)
 		}
 		converged[name] = b // add to map
 	}
 	return converged, nil
 }
 
-// EtcdAddHostnameConvergedWatcher adds a watcher with a callback that runs on
+// AddHostnameConvergedWatcher adds a watcher with a callback that runs on
 // hostname state changes.
-func EtcdAddHostnameConvergedWatcher(obj *EmbdEtcd, callbackFn func(map[string]bool) error) (func(), error) {
+func AddHostnameConvergedWatcher(obj *EmbdEtcd, callbackFn func(map[string]bool) error) (func(), error) {
 	path := fmt.Sprintf("/%s/converged/", NS)
 	internalCbFn := func(re *RE) error {
 		// TODO: get the value from the response, and apply delta...
 		// for now, just run a get operation which is easier to code!
-		m, err := EtcdHostnameConverged(obj)
+		m, err := HostnameConverged(obj)
 		if err != nil {
 			return err
 		}
@@ -1936,49 +1937,49 @@ func EtcdAddHostnameConvergedWatcher(obj *EmbdEtcd, callbackFn func(map[string]b
 	return obj.AddWatcher(path, internalCbFn, true, true, etcd.WithPrefix()) // no block and no converger reset
 }
 
-// EtcdSetClusterSize sets the ideal target cluster size of etcd peers
-func EtcdSetClusterSize(obj *EmbdEtcd, value uint16) error {
+// SetClusterSize sets the ideal target cluster size of etcd peers
+func SetClusterSize(obj *EmbdEtcd, value uint16) error {
 	if obj.flags.Trace {
-		log.Printf("Trace: Etcd: EtcdSetClusterSize(): %v", value)
-		defer log.Printf("Trace: Etcd: EtcdSetClusterSize(): Finished!")
+		log.Printf("Trace: Etcd: SetClusterSize(): %v", value)
+		defer log.Printf("Trace: Etcd: SetClusterSize(): Finished!")
 	}
 	key := fmt.Sprintf("/%s/idealClusterSize", NS)
 
 	if err := obj.Set(key, strconv.FormatUint(uint64(value), 10)); err != nil {
-		return fmt.Errorf("Etcd: SetClusterSize failed!") // exit in progress?
+		return fmt.Errorf("function SetClusterSize failed: %v", err) // exit in progress?
 	}
 	return nil
 }
 
-// EtcdGetClusterSize gets the ideal target cluster size of etcd peers
-func EtcdGetClusterSize(obj *EmbdEtcd) (uint16, error) {
+// GetClusterSize gets the ideal target cluster size of etcd peers
+func GetClusterSize(obj *EmbdEtcd) (uint16, error) {
 	key := fmt.Sprintf("/%s/idealClusterSize", NS)
 	keyMap, err := obj.Get(key)
 	if err != nil {
-		return 0, fmt.Errorf("Etcd: GetClusterSize failed: %v", err)
+		return 0, fmt.Errorf("function GetClusterSize failed: %v", err)
 	}
 
 	val, exists := keyMap[key]
 	if !exists || val == "" {
-		return 0, fmt.Errorf("Etcd: GetClusterSize failed: %v", err)
+		return 0, fmt.Errorf("function GetClusterSize failed: %v", err)
 	}
 
 	v, err := strconv.ParseUint(val, 10, 16)
 	if err != nil {
-		return 0, fmt.Errorf("Etcd: GetClusterSize failed: %v", err)
+		return 0, fmt.Errorf("function GetClusterSize failed: %v", err)
 	}
 	return uint16(v), nil
 }
 
-// EtcdMemberAdd adds a member to the cluster.
-func EtcdMemberAdd(obj *EmbdEtcd, peerURLs etcdtypes.URLs) (*etcd.MemberAddResponse, error) {
+// MemberAdd adds a member to the cluster.
+func MemberAdd(obj *EmbdEtcd, peerURLs etcdtypes.URLs) (*etcd.MemberAddResponse, error) {
 	//obj.Connect(false) // TODO: ?
 	ctx := context.Background()
 	var response *etcd.MemberAddResponse
 	var err error
 	for {
 		if obj.exiting { // the exit signal has been sent!
-			return nil, fmt.Errorf("Exiting...")
+			return nil, fmt.Errorf("exiting etcd")
 		}
 		obj.rLock.RLock()
 		response, err = obj.client.MemberAdd(ctx, peerURLs.StringSlice())
@@ -1993,15 +1994,15 @@ func EtcdMemberAdd(obj *EmbdEtcd, peerURLs etcdtypes.URLs) (*etcd.MemberAddRespo
 	return response, nil
 }
 
-// EtcdMemberRemove removes a member by mID and returns if it worked, and also
+// MemberRemove removes a member by mID and returns if it worked, and also
 // if there was an error. This is because it might have run without error, but
 // the member wasn't found, for example.
-func EtcdMemberRemove(obj *EmbdEtcd, mID uint64) (bool, error) {
+func MemberRemove(obj *EmbdEtcd, mID uint64) (bool, error) {
 	//obj.Connect(false) // TODO: ?
 	ctx := context.Background()
 	for {
 		if obj.exiting { // the exit signal has been sent!
-			return false, fmt.Errorf("Exiting...")
+			return false, fmt.Errorf("exiting etcd")
 		}
 		obj.rLock.RLock()
 		_, err := obj.client.MemberRemove(ctx, mID)
@@ -2019,21 +2020,21 @@ func EtcdMemberRemove(obj *EmbdEtcd, mID uint64) (bool, error) {
 	return true, nil
 }
 
-// EtcdMembers returns information on cluster membership.
+// Members returns information on cluster membership.
 // The member ID's are the keys, because an empty names means unstarted!
 // TODO: consider queueing this through the main loop with CtxError(ctx, err)
-func EtcdMembers(obj *EmbdEtcd) (map[uint64]string, error) {
+func Members(obj *EmbdEtcd) (map[uint64]string, error) {
 	//obj.Connect(false) // TODO: ?
 	ctx := context.Background()
 	var response *etcd.MemberListResponse
 	var err error
 	for {
 		if obj.exiting { // the exit signal has been sent!
-			return nil, fmt.Errorf("Exiting...")
+			return nil, fmt.Errorf("exiting etcd")
 		}
 		obj.rLock.RLock()
 		if obj.flags.Trace {
-			log.Printf("Trace: Etcd: EtcdMembers(): Endpoints are: %v", obj.client.Endpoints())
+			log.Printf("Trace: Etcd: Members(): Endpoints are: %v", obj.client.Endpoints())
 		}
 		response, err = obj.client.MemberList(ctx)
 		obj.rLock.RUnlock()
@@ -2052,18 +2053,18 @@ func EtcdMembers(obj *EmbdEtcd) (map[uint64]string, error) {
 	return members, nil
 }
 
-// EtcdLeader returns the current leader of the etcd server cluster
-func EtcdLeader(obj *EmbdEtcd) (string, error) {
+// Leader returns the current leader of the etcd server cluster
+func Leader(obj *EmbdEtcd) (string, error) {
 	//obj.Connect(false) // TODO: ?
 	var err error
 	membersMap := make(map[uint64]string)
-	if membersMap, err = EtcdMembers(obj); err != nil {
+	if membersMap, err = Members(obj); err != nil {
 		return "", err
 	}
 	addresses := obj.LocalhostClientURLs() // heuristic, but probably correct
 	if len(addresses) == 0 {
 		// probably a programming error...
-		return "", fmt.Errorf("Etcd: Leader: Programming error!")
+		return "", fmt.Errorf("programming error")
 	}
 	endpoint := addresses[0].Host // FIXME: arbitrarily picked the first one
 
@@ -2072,7 +2073,7 @@ func EtcdLeader(obj *EmbdEtcd) (string, error) {
 	var response *etcd.StatusResponse
 	for {
 		if obj.exiting { // the exit signal has been sent!
-			return "", fmt.Errorf("Exiting...")
+			return "", fmt.Errorf("exiting etcd")
 		}
 
 		obj.rLock.RLock()
@@ -2092,20 +2093,20 @@ func EtcdLeader(obj *EmbdEtcd) (string, error) {
 			return name, nil
 		}
 	}
-	return "", fmt.Errorf("Etcd: Members map is not current!") // not found
+	return "", fmt.Errorf("members map is not current") // not found
 }
 
-// EtcdWatch returns a channel that outputs a true bool when activity occurs
+// WatchAll returns a channel that outputs a true bool when activity occurs
 // TODO: Filter our watch (on the server side if possible) based on the
 // collection prefixes and filters that we care about...
-func EtcdWatch(obj *EmbdEtcd) chan bool {
+func WatchAll(obj *EmbdEtcd) chan bool {
 	ch := make(chan bool, 1) // buffer it so we can measure it
 	path := fmt.Sprintf("/%s/exported/", NS)
 	callback := func(re *RE) error {
 		// TODO: is this even needed? it used to happen on conn errors
 		log.Printf("Etcd: Watch: Path: %v", path) // event
 		if re == nil || re.response.Canceled {
-			return fmt.Errorf("Etcd: Watch is empty!") // will cause a CtxError+retry
+			return fmt.Errorf("watch is empty") // will cause a CtxError+retry
 		}
 		// we normally need to check if anything changed since the last
 		// event, since a set (export) with no changes still causes the
@@ -2125,15 +2126,15 @@ func EtcdWatch(obj *EmbdEtcd) chan bool {
 	return ch
 }
 
-// EtcdSetResources exports all of the resources which we pass in to etcd
-func EtcdSetResources(obj *EmbdEtcd, hostname string, resourceList []resources.Res) error {
+// SetResources exports all of the resources which we pass in to etcd
+func SetResources(obj *EmbdEtcd, hostname string, resourceList []resources.Res) error {
 	// key structure is /$NS/exported/$hostname/resources/$uid = $data
 
 	var kindFilter []string // empty to get from everyone
 	hostnameFilter := []string{hostname}
 	// this is not a race because we should only be reading keys which we
 	// set, and there should not be any contention with other hosts here!
-	originals, err := EtcdGetResources(obj, hostnameFilter, kindFilter)
+	originals, err := GetResources(obj, hostnameFilter, kindFilter)
 	if err != nil {
 		return err
 	}
@@ -2154,7 +2155,7 @@ func EtcdSetResources(obj *EmbdEtcd, hostname string, resourceList []resources.R
 			ifs = append(ifs, etcd.Compare(etcd.Value(path), "=", data)) // desired state
 			ops = append(ops, etcd.OpPut(path, data))
 		} else {
-			return fmt.Errorf("Etcd: SetResources: Error: Can't convert to B64: %v", err)
+			return fmt.Errorf("can't convert to B64: %v", err)
 		}
 	}
 
@@ -2196,18 +2197,18 @@ func EtcdSetResources(obj *EmbdEtcd, hostname string, resourceList []resources.R
 	return err
 }
 
-// EtcdGetResources collects all of the resources which match a filter from etcd
+// GetResources collects all of the resources which match a filter from etcd
 // If the kindfilter or hostnameFilter is empty, then it assumes no filtering...
 // TODO: Expand this with a more powerful filter based on what we eventually
 // support in our collect DSL. Ideally a server side filter like WithFilter()
 // We could do this if the pattern was /$NS/exported/$kind/$hostname/$uid = $data
-func EtcdGetResources(obj *EmbdEtcd, hostnameFilter, kindFilter []string) ([]resources.Res, error) {
+func GetResources(obj *EmbdEtcd, hostnameFilter, kindFilter []string) ([]resources.Res, error) {
 	// key structure is /$NS/exported/$hostname/resources/$uid = $data
 	path := fmt.Sprintf("/%s/exported/", NS)
 	resourceList := []resources.Res{}
 	keyMap, err := obj.Get(path, etcd.WithPrefix(), etcd.WithSort(etcd.SortByKey, etcd.SortAscend))
 	if err != nil {
-		return nil, fmt.Errorf("Etcd: GetResources: Error: Could not get resources: %v", err)
+		return nil, fmt.Errorf("could not get resources: %v", err)
 	}
 	for key, val := range keyMap {
 		if !strings.HasPrefix(key, path) { // sanity check
@@ -2216,14 +2217,14 @@ func EtcdGetResources(obj *EmbdEtcd, hostnameFilter, kindFilter []string) ([]res
 
 		str := strings.Split(key[len(path):], "/")
 		if len(str) != 4 {
-			return nil, fmt.Errorf("Etcd: GetResources: Error: Unexpected chunk count!")
+			return nil, fmt.Errorf("unexpected chunk count")
 		}
 		hostname, r, kind, name := str[0], str[1], str[2], str[3]
 		if r != "resources" {
-			return nil, fmt.Errorf("Etcd: GetResources: Error: Unexpected chunk pattern!")
+			return nil, fmt.Errorf("unexpected chunk pattern")
 		}
 		if kind == "" {
-			return nil, fmt.Errorf("Etcd: GetResources: Error: Unexpected kind chunk!")
+			return nil, fmt.Errorf("unexpected kind chunk")
 		}
 
 		// FIXME: ideally this would be a server side filter instead!
@@ -2241,7 +2242,7 @@ func EtcdGetResources(obj *EmbdEtcd, hostnameFilter, kindFilter []string) ([]res
 			log.Printf("Etcd: Get: (Hostname, Kind, Name): (%s, %s, %s)", hostname, kind, name)
 			resourceList = append(resourceList, obj)
 		} else {
-			return nil, fmt.Errorf("Etcd: GetResources: Error: Can't convert from B64: %v", err)
+			return nil, fmt.Errorf("can't convert from B64: %v", err)
 		}
 	}
 	return resourceList, nil
@@ -2269,11 +2270,11 @@ func ApplyDeltaEvents(re *RE, urlsmap etcdtypes.URLsMap) (etcdtypes.URLsMap, err
 		case etcd.EventTypePut:
 			val := bytes.NewBuffer(event.Kv.Value).String()
 			if val == "" {
-				return nil, fmt.Errorf("Etcd: ApplyDeltaEvents: Value is empty!")
+				return nil, fmt.Errorf("value in ApplyDeltaEvents is empty")
 			}
 			urls, err := etcdtypes.NewURLs(strings.Split(val, ","))
 			if err != nil {
-				return nil, fmt.Errorf("Etcd: ApplyDeltaEvents: Format error: %v", err)
+				return nil, fmt.Errorf("format error in ApplyDeltaEvents: %v", err)
 			}
 			urlsmap[key] = urls // add to map
 
@@ -2290,7 +2291,7 @@ func ApplyDeltaEvents(re *RE, urlsmap etcdtypes.URLsMap) (etcdtypes.URLsMap, err
 			delete(urlsmap, key)
 
 		default:
-			return nil, fmt.Errorf("Etcd: ApplyDeltaEvents: Error: Unknown event: %+v", event.Type)
+			return nil, fmt.Errorf("unknown event in ApplyDeltaEvents: %+v", event.Type)
 		}
 	}
 	return urlsmap, nil

etcd/str.go

@@ -0,0 +1,115 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package etcd
+
+import (
+	"fmt"
+	"strings"
+
+	"github.com/purpleidea/mgmt/util"
+
+	etcd "github.com/coreos/etcd/clientv3"
+	errwrap "github.com/pkg/errors"
+)
+
+// WatchStr returns a channel which spits out events on key activity.
+// FIXME: It should close the channel when it's done, and spit out errors when
+// something goes wrong.
+func WatchStr(obj *EmbdEtcd, key string) chan error {
+	// new key structure is /$NS/strings/$key/$hostname = $data
+	path := fmt.Sprintf("/%s/strings/%s", NS, key)
+	ch := make(chan error, 1)
+	// FIXME: fix our API so that we get a close event on shutdown.
+	callback := func(re *RE) error {
+		// TODO: is this even needed? it used to happen on conn errors
+		//log.Printf("Etcd: Watch: Path: %v", path) // event
+		if re == nil || re.response.Canceled {
+			return fmt.Errorf("watch is empty") // will cause a CtxError+retry
+		}
+		if len(ch) == 0 { // send event only if one isn't pending
+			ch <- nil // event
+		}
+		return nil
+	}
+	_, _ = obj.AddWatcher(path, callback, true, false, etcd.WithPrefix()) // no need to check errors
+	return ch
+}
+
+// GetStr collects all of the strings which match a namespace in etcd.
+func GetStr(obj *EmbdEtcd, hostnameFilter []string, key string) (map[string]string, error) {
+	// old key structure is /$NS/strings/$hostname/$key = $data
+	// new key structure is /$NS/strings/$key/$hostname = $data
+	// FIXME: if we have the $key as the last token (old key structure), we
+	// can allow the key to contain the slash char, otherwise we need to
+	// verify that one isn't present in the input string.
+	path := fmt.Sprintf("/%s/strings/%s", NS, key)
+	keyMap, err := obj.Get(path, etcd.WithPrefix(), etcd.WithSort(etcd.SortByKey, etcd.SortAscend))
+	if err != nil {
+		return nil, errwrap.Wrapf(err, "could not get strings in: %s", key)
+	}
+	result := make(map[string]string)
+	for key, val := range keyMap {
+		if !strings.HasPrefix(key, path) { // sanity check
+			continue
+		}
+
+		str := strings.Split(key[len(path):], "/")
+		if len(str) != 2 {
+			return nil, fmt.Errorf("unexpected chunk count of %d", len(str))
+		}
+		_, hostname := str[0], str[1]
+
+		if hostname == "" {
+			return nil, fmt.Errorf("unexpected chunk length of %d", len(hostname))
+		}
+
+		// FIXME: ideally this would be a server side filter instead!
+		if len(hostnameFilter) > 0 && !util.StrInList(hostname, hostnameFilter) {
+			continue
+		}
+		//log.Printf("Etcd: GetStr(%s): (Hostname, Data): (%s, %s)", key, hostname, val)
+		result[hostname] = val
+	}
+	return result, nil
+}
+
+// SetStr sets a key and hostname pair to a certain value. If the value is nil,
+// then it deletes the key. Otherwise the value should point to a string.
+// TODO: TTL or delete disconnect?
+func SetStr(obj *EmbdEtcd, hostname, key string, data *string) error {
+	// key structure is /$NS/strings/$key/$hostname = $data
+	path := fmt.Sprintf("/%s/strings/%s/%s", NS, key, hostname)
+	ifs := []etcd.Cmp{} // list matching the desired state
+	ops := []etcd.Op{}  // list of ops in this transaction (then)
+	els := []etcd.Op{}  // list of ops in this transaction (else)
+	if data == nil {    // perform a delete
+		// TODO: use https://github.com/coreos/etcd/pull/7417 if merged
+		//ifs = append(ifs, etcd.KeyExists(path))
+		ifs = append(ifs, etcd.Compare(etcd.Version(path), ">", 0))
+		ops = append(ops, etcd.OpDelete(path))
+	} else {
+		data := *data                                                // get the real value
+		ifs = append(ifs, etcd.Compare(etcd.Value(path), "=", data)) // desired state
+		els = append(els, etcd.OpPut(path, data))
+	}
+
+	// it's important to do this in one transaction, and atomically, because
+	// this way, we only generate one watch event, and only when it's needed
+	_, err := obj.Txn(ifs, ops, els) // TODO: do we need to look at response?
+	return errwrap.Wrapf(err, "could not set strings in: %s", key)
+}

etcd/world.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -30,7 +30,7 @@ type World struct {
 // ResExport exports a list of resources under our hostname namespace.
 // Subsequent calls replace the previously set collection atomically.
 func (obj *World) ResExport(resourceList []resources.Res) error {
-	return EtcdSetResources(obj.EmbdEtcd, obj.Hostname, resourceList)
+	return SetResources(obj.EmbdEtcd, obj.Hostname, resourceList)
 }
 
 // ResCollect gets the collection of exported resources which match the filter.
@@ -39,5 +39,26 @@ func (obj *World) ResCollect(hostnameFilter, kindFilter []string) ([]resources.R
 	// XXX: should we be restricted to retrieving resources that were
 	// exported with a tag that allows or restricts our hostname? We could
 	// enforce that here if the underlying API supported it... Add this?
-	return EtcdGetResources(obj.EmbdEtcd, hostnameFilter, kindFilter)
+	return GetResources(obj.EmbdEtcd, hostnameFilter, kindFilter)
+}
+
+// SetWatch returns a channel which spits out events on possible string changes.
+func (obj *World) StrWatch(namespace string) chan error {
+	return WatchStr(obj.EmbdEtcd, namespace)
+}
+
+// StrGet returns a map of hostnames to values in the given namespace.
+func (obj *World) StrGet(namespace string) (map[string]string, error) {
+	return GetStr(obj.EmbdEtcd, []string{}, namespace)
+}
+
+// StrSet sets the namespace value to a particular string under the identity of
+// its own hostname.
+func (obj *World) StrSet(namespace, value string) error {
+	return SetStr(obj.EmbdEtcd, obj.Hostname, namespace, &value)
+}
+
+// StrDel deletes the value in a particular namespace.
+func (obj *World) StrDel(namespace string) error {
+	return SetStr(obj.EmbdEtcd, obj.Hostname, namespace, nil)
 }

event/event.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -22,14 +22,14 @@ import (
 	"fmt"
 )
 
-//go:generate stringer -type=EventName -output=eventname_stringer.go
+//go:generate stringer -type=Kind -output=kind_stringer.go
 
-// EventName represents the type of event being passed.
-type EventName int
+// Kind represents the type of event being passed.
+type Kind int
 
-// The different event names are used in different contexts.
+// The different event kinds are used in different contexts.
 const (
-	EventNil EventName = iota
+	EventNil Kind = iota
 	EventExit
 	EventStart
 	EventPause
@@ -43,11 +43,10 @@ type Resp chan error
 
 // Event is the main struct that stores event information and responses.
 type Event struct {
-	Name EventName
+	Kind Kind
 	Resp Resp // channel to send an ack response on, nil to skip
 	//Wg   *sync.WaitGroup // receiver barrier to Wait() for everyone else on
-	Msg      string // some words for fun
-	Activity bool   // did something interesting happen?
+	Err error // store an error in our event
 }
 
 // ACK sends a single acknowledgement on the channel if one was requested.
@@ -80,7 +79,7 @@ func NewResp() Resp {
 // ACK sends a true value to resp.
 func (resp Resp) ACK() {
 	if resp != nil {
-		resp <- nil
+		resp <- nil // TODO: close instead?
 	}
 }
 
@@ -114,7 +113,7 @@ func (resp Resp) ACKWait() {
 	}
 }
 
-// GetActivity returns the activity value.
-func (event *Event) GetActivity() bool {
-	return event.Activity
+// Error returns the stored error value.
+func (event *Event) Error() error {
+	return event.Err
 }

examples/augeas1.yaml

@@ -0,0 +1,11 @@
+---
+graph: mygraph
+resources:
+  augeas:
+  - name: sshd_config
+    lens: Sshd.lns
+    file: "/etc/ssh/sshd_config"
+    sets:
+    - path: X11Forwarding
+      value: false
+edges:

examples/autoedges3.yaml

@@ -5,11 +5,13 @@ resources:
   - name: drbd-utils
     meta:
       autoedge: true
+      noop: true
     state: installed
   file:
   - name: file1
     meta:
       autoedge: true
+      noop: true
     path: "/etc/drbd.conf"
     content: |
       # this is an mgmt test
@@ -17,13 +19,14 @@ resources:
   - name: file2
     meta:
       autoedge: true
+      noop: true
     path: "/etc/drbd.d/"
-    content: |
-      i am a directory
+    source: /dev/null
     state: exists
   svc:
   - name: drbd
     meta:
       autoedge: true
+      noop: true
     state: stopped
 edges: []

examples/deep-dirs.yaml

@@ -0,0 +1,9 @@
+---
+graph: mygraph
+resources:
+  file:
+  - name: file1
+    path: "/tmp/mgmt/a/b/c/f1"
+    content: |
+      i am f1
+    state: exists

examples/etcd1a.yaml

@@ -2,15 +2,10 @@
 graph: mygraph
 resources:
   file:
-  - name: file1a
-    path: "/tmp/mgmtA/f1a"
+  - name: "@@filea"
+    path: "/tmp/mgmtA/fA"
     content: |
-      i am f1
-    state: exists
-  - name: "@@file2a"
-    path: "/tmp/mgmtA/f2a"
-    content: |
-      i am f2, exported from host A
+      i am fA, exported from host A
     state: exists
 collect:
 - kind: file

examples/etcd1b.yaml

@@ -2,15 +2,10 @@
 graph: mygraph
 resources:
   file:
-  - name: file1b
-    path: "/tmp/mgmtB/f1b"
+  - name: "@@fileb"
+    path: "/tmp/mgmtB/fB"
     content: |
-      i am f1
-    state: exists
-  - name: "@@file2b"
-    path: "/tmp/mgmtB/f2b"
-    content: |
-      i am f2, exported from host B
+      i am fB, exported from host B
     state: exists
 collect:
 - kind: file

examples/etcd1c.yaml

@@ -2,15 +2,10 @@
 graph: mygraph
 resources:
   file:
-  - name: file1c
-    path: "/tmp/mgmtC/f1c"
+  - name: "@@filec"
+    path: "/tmp/mgmtC/fC"
     content: |
-      i am f1
-    state: exists
-  - name: "@@file2c"
-    path: "/tmp/mgmtC/f2c"
-    content: |
-      i am f2, exported from host C
+      i am fC, exported from host C
     state: exists
 collect:
 - kind: file

examples/etcd1d.yaml

@@ -2,15 +2,10 @@
 graph: mygraph
 resources:
   file:
-  - name: file1d
-    path: "/tmp/mgmtD/f1d"
+  - name: "@@filed"
+    path: "/tmp/mgmtD/fD"
     content: |
-      i am f1
-    state: exists
-  - name: "@@file2d"
-    path: "/tmp/mgmtD/f2d"
-    content: |
-      i am f2, exported from host D
+      i am fD, exported from host D
     state: exists
 collect:
 - kind: file

examples/etcd1e.yaml

@@ -0,0 +1,13 @@
+---
+graph: mygraph
+resources:
+  file:
+  - name: "@@filee"
+    path: "/tmp/mgmtE/fE"
+    content: |
+      i am fE, exported from host E
+    state: exists
+collect:
+- kind: file
+  pattern: "/tmp/mgmtE/"
+edges: []

examples/exec3-sema.yaml

@@ -0,0 +1,67 @@
+---
+graph: parallel
+resources:
+  exec:
+  - name: pkg10
+    meta:
+      sema: ['mylock:1', 'otherlock:42']
+    cmd: sleep 10s
+    shell: ''
+    timeout: 0
+    watchcmd: ''
+    watchshell: ''
+    ifcmd: ''
+    ifshell: ''
+    pollint: 0
+    state: present
+  - name: svc10
+    meta:
+      sema: ['mylock:1']
+    cmd: sleep 10s
+    shell: ''
+    timeout: 0
+    watchcmd: ''
+    watchshell: ''
+    ifcmd: ''
+    ifshell: ''
+    pollint: 0
+    state: present
+  - name: exec10
+    meta:
+      sema: ['mylock:1']
+    cmd: sleep 10s
+    shell: ''
+    timeout: 0
+    watchcmd: ''
+    watchshell: ''
+    ifcmd: ''
+    ifshell: ''
+    pollint: 0
+    state: present
+  - name: pkg15
+    meta:
+      sema: ['mylock:1', 'otherlock:42']
+    cmd: sleep 15s
+    shell: ''
+    timeout: 0
+    watchcmd: ''
+    watchshell: ''
+    ifcmd: ''
+    ifshell: ''
+    pollint: 0
+    state: present
+edges:
+- name: e1
+  from:
+    kind: exec
+    name: pkg10
+  to:
+    kind: exec
+    name: svc10
+- name: e2
+  from:
+    kind: exec
+    name: svc10
+  to:
+    kind: exec
+    name: exec10

examples/file3.yaml

@@ -1,14 +1,13 @@
 ---
 graph: mygraph
-comment: You can test Watch and CheckApply failures with chmod ugo-r and chmod ugo-w.
 resources:
   file:
   - name: file1
-    path: "/tmp/mgmt/f1"
     meta:
-      retry: 3
-      delay: 5000
+      limit: .inf
+      burst: 0
+    path: "/tmp/mgmt/hello"
     content: |
-      i am f1
+      i am a file
     state: exists
 edges: []

examples/file4.yaml

@@ -0,0 +1,10 @@
+---
+graph: mygraph
+resources:
+  file:
+  - name: file1
+    path: "/tmp/mgmt/hello"
+    content: |
+      i am a file
+    state: exists
+edges: []

examples/kv1.yaml

@@ -0,0 +1,8 @@
+---
+graph: mygraph
+resources:
+  kv:
+  - name: kv1
+    key: "hello"
+    value: "world"
+edges: []

examples/kv2.yaml

@@ -0,0 +1,7 @@
+---
+graph: mygraph
+resources:
+  kv:
+  - name: kv1
+    key: "iamdeleted"
+edges: []

examples/kv3.yaml

@@ -0,0 +1,9 @@
+---
+graph: mygraph
+resources:
+  kv:
+  - name: kv1
+    key: "stage"
+    value: "3"
+    skiplessthan: true
+edges: []

examples/kv4.yaml

@@ -0,0 +1,31 @@
+---
+graph: mygraph
+resources:
+  kv:
+  - name: kv1
+    key: "stage"
+    value: "1"
+    skiplessthan: true
+  - name: kv2
+    key: "stage"
+    value: "2"
+    skiplessthan: true
+  - name: kv3
+    key: "stage"
+    value: "3"
+    skiplessthan: true
+edges:
+- name: e1
+  from:
+    kind: kv
+    name: kv1
+  to:
+    kind: kv
+    name: kv2
+- name: e2
+  from:
+    kind: kv
+    name: kv2
+  to:
+    kind: kv
+    name: kv3

examples/lib/libmgmt1.go

@@ -40,10 +40,10 @@ func NewMyGAPI(data gapi.Data, name string, interval uint) (*MyGAPI, error) {
 // Init initializes the MyGAPI struct.
 func (obj *MyGAPI) Init(data gapi.Data) error {
 	if obj.initialized {
-		return fmt.Errorf("Already initialized!")
+		return fmt.Errorf("already initialized")
 	}
 	if obj.Name == "" {
-		return fmt.Errorf("The graph name must be specified!")
+		return fmt.Errorf("the graph name must be specified")
 	}
 	obj.data = data // store for later
 	obj.closeChan = make(chan struct{})
@@ -59,7 +59,7 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 
 	n1, err := resources.NewNoopRes("noop1")
 	if err != nil {
-		return nil, fmt.Errorf("Can't create resource: %v", err)
+		return nil, fmt.Errorf("can't create resource: %v", err)
 	}
 
 	// we can still build a graph via the yaml method
@@ -85,8 +85,8 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 	return g, err
 }
 
-// SwitchStream returns nil errors every time there could be a new graph.
-func (obj *MyGAPI) SwitchStream() chan error {
+// Next returns nil errors every time there could be a new graph.
+func (obj *MyGAPI) Next() chan error {
 	if obj.data.NoWatch || obj.Interval <= 0 {
 		return nil
 	}
@@ -107,7 +107,11 @@ func (obj *MyGAPI) SwitchStream() chan error {
 			select {
 			case <-ticker.C:
 				log.Printf("libmgmt: Generating new graph...")
-				ch <- nil // trigger a run
+				select {
+				case ch <- nil: // trigger a run
+				case <-obj.closeChan:
+					return
+				}
 			case <-obj.closeChan:
 				return
 			}
@@ -164,7 +168,7 @@ func Run() error {
 				return
 			}
 			log.Println("Interrupted by signal")
-			obj.Exit(fmt.Errorf("Killed by %v", sig))
+			obj.Exit(fmt.Errorf("killed by %v", sig))
 			return
 		case <-exit:
 			return

examples/lib/libmgmt2.go

@@ -42,10 +42,10 @@ func NewMyGAPI(data gapi.Data, name string, interval uint, count uint) (*MyGAPI,
 // Init initializes the MyGAPI struct.
 func (obj *MyGAPI) Init(data gapi.Data) error {
 	if obj.initialized {
-		return fmt.Errorf("Already initialized!")
+		return fmt.Errorf("already initialized")
 	}
 	if obj.Name == "" {
-		return fmt.Errorf("The graph name must be specified!")
+		return fmt.Errorf("the graph name must be specified")
 	}
 	obj.data = data // store for later
 	obj.closeChan = make(chan struct{})
@@ -64,7 +64,7 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 	for i := uint(0); i < obj.Count; i++ {
 		n, err := resources.NewNoopRes(fmt.Sprintf("noop%d", i))
 		if err != nil {
-			return nil, fmt.Errorf("Can't create resource: %v", err)
+			return nil, fmt.Errorf("can't create resource: %v", err)
 		}
 		v := pgraph.NewVertex(n)
 		g.AddVertex(v)
@@ -78,8 +78,8 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 	return g, nil
 }
 
-// SwitchStream returns nil errors every time there could be a new graph.
-func (obj *MyGAPI) SwitchStream() chan error {
+// Next returns nil errors every time there could be a new graph.
+func (obj *MyGAPI) Next() chan error {
 	if obj.data.NoWatch || obj.Interval <= 0 {
 		return nil
 	}
@@ -100,7 +100,11 @@ func (obj *MyGAPI) SwitchStream() chan error {
 			select {
 			case <-ticker.C:
 				log.Printf("libmgmt: Generating new graph...")
-				ch <- nil // trigger a run
+				select {
+				case ch <- nil: // trigger a run
+				case <-obj.closeChan:
+					return
+				}
 			case <-obj.closeChan:
 				return
 			}
@@ -158,7 +162,7 @@ func Run(count uint) error {
 				return
 			}
 			log.Println("Interrupted by signal")
-			obj.Exit(fmt.Errorf("Killed by %v", sig))
+			obj.Exit(fmt.Errorf("killed by %v", sig))
 			return
 		case <-exit:
 			return

examples/lib/libmgmt3.go

@@ -14,6 +14,8 @@ import (
 	mgmt "github.com/purpleidea/mgmt/lib"
 	"github.com/purpleidea/mgmt/pgraph"
 	"github.com/purpleidea/mgmt/resources"
+
+	"golang.org/x/time/rate"
 )
 
 // MyGAPI implements the main GAPI interface.
@@ -39,10 +41,10 @@ func NewMyGAPI(data gapi.Data, name string, interval uint) (*MyGAPI, error) {
 // Init initializes the MyGAPI struct.
 func (obj *MyGAPI) Init(data gapi.Data) error {
 	if obj.initialized {
-		return fmt.Errorf("Already initialized!")
+		return fmt.Errorf("already initialized")
 	}
 	if obj.Name == "" {
-		return fmt.Errorf("The graph name must be specified!")
+		return fmt.Errorf("the graph name must be specified")
 	}
 	obj.data = data // store for later
 	obj.closeChan = make(chan struct{})
@@ -58,10 +60,17 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 
 	g := pgraph.NewGraph(obj.Name)
 
+	// FIXME: these are being specified temporarily until it's the default!
+	metaparams := resources.MetaParams{
+		Limit: rate.Inf,
+		Burst: 0,
+	}
+
 	content := "Delete me to trigger a notification!\n"
 	f0 := &resources.FileRes{
 		BaseRes: resources.BaseRes{
-			Name: "README",
+			Name:       "README",
+			MetaParams: metaparams,
 		},
 		Path:    "/tmp/mgmt/README",
 		Content: &content,
@@ -73,7 +82,8 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 
 	p1 := &resources.PasswordRes{
 		BaseRes: resources.BaseRes{
-			Name: "password1",
+			Name:       "password1",
+			MetaParams: metaparams,
 		},
 		Length: 8,    // generated string will have this many characters
 		Saved:  true, // this causes passwords to be stored in plain text!
@@ -83,10 +93,11 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 
 	f1 := &resources.FileRes{
 		BaseRes: resources.BaseRes{
-			Name: "file1",
+			Name:       "file1",
+			MetaParams: metaparams,
 			// send->recv!
 			Recv: map[string]*resources.Send{
-				"Content": &resources.Send{Res: p1, Key: "Password"},
+				"Content": {Res: p1, Key: "Password"},
 			},
 		},
 		Path: "/tmp/mgmt/secret",
@@ -99,7 +110,8 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 
 	n1 := &resources.NoopRes{
 		BaseRes: resources.BaseRes{
-			Name: "noop1",
+			Name:       "noop1",
+			MetaParams: metaparams,
 		},
 	}
 
@@ -120,8 +132,8 @@ func (obj *MyGAPI) Graph() (*pgraph.Graph, error) {
 	return g, nil
 }
 
-// SwitchStream returns nil errors every time there could be a new graph.
-func (obj *MyGAPI) SwitchStream() chan error {
+// Next returns nil errors every time there could be a new graph.
+func (obj *MyGAPI) Next() chan error {
 	if obj.data.NoWatch || obj.Interval <= 0 {
 		return nil
 	}
@@ -142,7 +154,11 @@ func (obj *MyGAPI) SwitchStream() chan error {
 			select {
 			case <-ticker.C:
 				log.Printf("libmgmt: Generating new graph...")
-				ch <- nil // trigger a run
+				select {
+				case ch <- nil: // trigger a run
+				case <-obj.closeChan:
+					return
+				}
 			case <-obj.closeChan:
 				return
 			}
@@ -201,7 +217,7 @@ func Run() error {
 				return
 			}
 			log.Println("Interrupted by signal")
-			obj.Exit(fmt.Errorf("Killed by %v", sig))
+			obj.Exit(fmt.Errorf("killed by %v", sig))
 			return
 		case <-exit:
 			return

examples/limit1.yaml

@@ -0,0 +1,13 @@
+---
+graph: mygraph
+resources:
+  file:
+  - name: file1
+    meta:
+      limit: 0.2
+      burst: 5
+    path: "/tmp/mgmt/limit"
+    content: |
+      i am a normal file
+    state: exists
+edges: []

examples/noop2.yaml

@@ -0,0 +1,30 @@
+---
+graph: mygraph
+comment: dangerous noop example
+resources:
+  noop:
+  - name: noop1
+    meta:
+      noop: true
+  file:
+  - name: file1
+    path: "/tmp/mgmt/hello-noop"
+    content: |
+      hello world from @purpleidea
+    state: exists
+    meta:
+      noop: true
+  exec:
+  - name: exec1
+    meta:
+      noop: true
+    cmd: 'rm -rf /'
+    shell: '/bin/bash'
+    timeout: 0
+    watchcmd: ''
+    watchshell: ''
+    ifcmd: ''
+    ifshell: ''
+    pollint: 0
+    state: present
+edges: []

examples/poll1.yaml

@@ -0,0 +1,24 @@
+---
+graph: mygraph
+resources:
+  file:
+  - name: file1
+    meta:
+      poll: 5
+    path: "/tmp/mgmt/f1"
+    content: |
+      i poll every 5 seconds
+    state: exists
+  - name: file2
+    path: "/tmp/mgmt/f2"
+    content: |
+      i use the event based watcher
+    state: exists
+  - name: file3
+    meta:
+      poll: 1
+    path: "/tmp/mgmt/f3"
+    content: |
+      i poll every second
+    state: exists
+edges: []

examples/retry1.yaml

@@ -0,0 +1,57 @@
+---
+graph: mygraph
+comment: You can test Watch and CheckApply failures with chmod ugo-r and chmod ugo-w.
+resources:
+  exec:
+  - name: exec1
+    cmd: 'touch /tmp/mgmt/no-read && chmod ugo-r /tmp/mgmt/no-read'
+    shell: '/bin/bash'
+    timeout: 0
+    watchcmd: ''
+    watchshell: ''
+    ifcmd: ''
+    ifshell: ''
+    pollint: 0
+    state: present
+  - name: exec2
+    cmd: 'touch /tmp/mgmt/no-write && chmod ugo-w /tmp/mgmt/no-write'
+    shell: '/bin/bash'
+    timeout: 0
+    watchcmd: ''
+    watchshell: ''
+    ifcmd: ''
+    ifshell: ''
+    pollint: 0
+    state: present
+  file:
+  - name: noread
+    path: "/tmp/mgmt/no-read"
+    meta:
+      retry: 3
+      delay: 5000
+    content: |
+      i am f1
+    state: exists
+  - name: nowrite
+    path: "/tmp/mgmt/no-write"
+    meta:
+      retry: 3
+      delay: 5000
+    content: |
+      i am f1
+    state: exists
+edges:
+- name: e1
+  from:
+    kind: exec
+    name: exec1
+  to:
+    kind: file
+    name: noread
+- name: e2
+  from:
+    kind: exec
+    name: exec2
+  to:
+    kind: file
+    name: nowrite

examples/timer1.yaml

@@ -4,7 +4,7 @@ comment: timer example
 resources:
   timer:
   - name: timer1
-    interval: 30
+    interval: 3
   exec:
   - name: exec1
     cmd: echo hello world

examples/virt4.yaml

@@ -0,0 +1,21 @@
+---
+graph: mygraph
+resources:
+  virt:
+  - name: mgmt4
+    meta:
+      limit: .inf
+      burst: 0
+    uri: 'qemu:///session'
+    cpus: 1
+    maxcpus: 4
+    memory: 524288
+    boot:
+    - hd
+    disk:
+    - type: qcow2
+      source: "~/.local/share/libvirt/images/fedora-23-scratch.qcow2"
+    state: running
+    transient: false
+edges: []
+comment: "qemu-img create -b fedora-23.qcow2 -f qcow2 fedora-23-scratch.qcow2"

gapi/gapi.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -23,19 +23,10 @@ import (
 	"github.com/purpleidea/mgmt/resources"
 )
 
-// World is an interface to the rest of the different graph state. It allows
-// the GAPI to store state and exchange information throughout the cluster. It
-// is the interface each machine uses to communicate with the rest of the world.
-type World interface { // TODO: is there a better name for this interface?
-	ResExport([]resources.Res) error
-	// FIXME: should this method take a "filter" data struct instead of many args?
-	ResCollect(hostnameFilter, kindFilter []string) ([]resources.Res, error)
-}
-
 // Data is the set of input values passed into the GAPI structs via Init.
 type Data struct {
 	Hostname string // uuid for the host, required for GAPI
-	World    World
+	World    resources.World
 	Noop     bool
 	NoWatch  bool
 	// NOTE: we can add more fields here if needed by GAPI endpoints
@@ -45,6 +36,6 @@ type Data struct {
 type GAPI interface {
 	Init(Data) error               // initializes the GAPI and passes in useful data
 	Graph() (*pgraph.Graph, error) // returns the most recent pgraph
-	SwitchStream() chan error      // returns a stream of switch events
+	Next() chan error              // returns a stream of switch events
 	Close() error                  // shutdown the GAPI
 }

lib/cli.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -55,17 +55,17 @@ func run(c *cli.Context) error {
 
 	if _ = c.String("code"); c.IsSet("code") {
 		if obj.GAPI != nil {
-			return fmt.Errorf("Can't combine code GAPI with existing GAPI.")
+			return fmt.Errorf("can't combine code GAPI with existing GAPI")
 		}
 		// TODO: implement DSL GAPI
 		//obj.GAPI = &dsl.GAPI{
 		//	Code: &s,
 		//}
-		return fmt.Errorf("The Code GAPI is not implemented yet!") // TODO: DSL
+		return fmt.Errorf("the Code GAPI is not implemented yet") // TODO: DSL
 	}
 	if y := c.String("yaml"); c.IsSet("yaml") {
 		if obj.GAPI != nil {
-			return fmt.Errorf("Can't combine YAML GAPI with existing GAPI.")
+			return fmt.Errorf("can't combine YAML GAPI with existing GAPI")
 		}
 		obj.GAPI = &yamlgraph.GAPI{
 			File: &y,
@@ -73,7 +73,7 @@ func run(c *cli.Context) error {
 	}
 	if p := c.String("puppet"); c.IsSet("puppet") {
 		if obj.GAPI != nil {
-			return fmt.Errorf("Can't combine puppet GAPI with existing GAPI.")
+			return fmt.Errorf("can't combine puppet GAPI with existing GAPI")
 		}
 		obj.GAPI = &puppet.GAPI{
 			PuppetParam: &p,
@@ -84,6 +84,7 @@ func run(c *cli.Context) error {
 
 	obj.NoWatch = c.Bool("no-watch")
 	obj.Noop = c.Bool("noop")
+	obj.Sema = c.Int("sema")
 	obj.Graphviz = c.String("graphviz")
 	obj.GraphvizFilter = c.String("graphviz-filter")
 	obj.ConvergedTimeout = c.Int("converged-timeout")
@@ -115,6 +116,9 @@ func run(c *cli.Context) error {
 		return err
 	}
 
+	obj.Prometheus = c.Bool("prometheus")
+	obj.PrometheusListen = c.String("prometheus-listen")
+
 	// install the exit signal handler
 	exit := make(chan struct{})
 	defer close(exit)
@@ -132,7 +136,7 @@ func run(c *cli.Context) error {
 				return
 			}
 			log.Println("Interrupted by signal")
-			obj.Exit(fmt.Errorf("Killed by %v", sig))
+			obj.Exit(fmt.Errorf("killed by %v", sig))
 			return
 		case <-exit:
 			return
@@ -152,7 +156,7 @@ func CLI(program, version string, flags Flags) error {
 
 	// test for sanity
 	if program == "" || version == "" {
-		return fmt.Errorf("Program was not compiled correctly. Please see Makefile.")
+		return fmt.Errorf("program was not compiled correctly, see Makefile")
 	}
 	app := cli.NewApp()
 	app.Name = program // App.name and App.version pass these values through
@@ -225,6 +229,11 @@ func CLI(program, version string, flags Flags) error {
 					Name:  "noop",
 					Usage: "globally force all resources into no-op mode",
 				},
+				cli.IntFlag{
+					Name:  "sema",
+					Value: -1,
+					Usage: "globally add a semaphore to all resources with this lock count",
+				},
 				cli.StringFlag{
 					Name:  "graphviz, g",
 					Value: "",
@@ -232,7 +241,7 @@ func CLI(program, version string, flags Flags) error {
 				},
 				cli.StringFlag{
 					Name:  "graphviz-filter, gf",
-					Value: "dot", // directed graph default
+					Value: "",
 					Usage: "graphviz filter to use",
 				},
 				cli.IntFlag{
@@ -320,6 +329,15 @@ func CLI(program, version string, flags Flags) error {
 					Value: "",
 					Usage: "default identity used for generation",
 				},
+				cli.BoolFlag{
+					Name:  "prometheus",
+					Usage: "start a prometheus instance",
+				},
+				cli.StringFlag{
+					Name:  "prometheus-listen",
+					Value: "",
+					Usage: "specify prometheus instance binding",
+				},
 			},
 		},
 	}

lib/main.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -23,7 +23,6 @@ import (
 	"log"
 	"os"
 	"path"
-	"sync"
 	"time"
 
 	"github.com/purpleidea/mgmt/converger"
@@ -31,6 +30,7 @@ import (
 	"github.com/purpleidea/mgmt/gapi"
 	"github.com/purpleidea/mgmt/pgp"
 	"github.com/purpleidea/mgmt/pgraph"
+	"github.com/purpleidea/mgmt/prometheus"
 	"github.com/purpleidea/mgmt/recwatch"
 	"github.com/purpleidea/mgmt/remote"
 	"github.com/purpleidea/mgmt/resources"
@@ -67,6 +67,7 @@ type Main struct {
 
 	NoWatch          bool   // do not update graph on watched graph definition file changes
 	Noop             bool   // globally force all resources into no-op mode
+	Sema             int    // add a semaphore with this lock count to each resource
 	Graphviz         string // output file for graphviz data
 	GraphvizFilter   string // graphviz filter to use
 	ConvergedTimeout int    // exit after approximately this many seconds in a converged state; -1 to disable
@@ -94,6 +95,9 @@ type Main struct {
 	PgpIdentity *string
 	pgpKeys     *pgp.PGP // agent key pair
 
+	Prometheus       bool   // enable prometheus metrics
+	PrometheusListen string // prometheus instance bind specification
+
 	exit chan error // exit signal
 }
 
@@ -101,11 +105,11 @@ type Main struct {
 func (obj *Main) Init() error {
 
 	if obj.Program == "" || obj.Version == "" {
-		return fmt.Errorf("You must set the Program and Version strings!")
+		return fmt.Errorf("you must set the Program and Version strings")
 	}
 
 	if obj.Prefix != nil && obj.TmpPrefix {
-		return fmt.Errorf("Choosing a prefix and the request for a tmp prefix is illogical!")
+		return fmt.Errorf("choosing a prefix and the request for a tmp prefix is illogical")
 	}
 
 	obj.idealClusterSize = uint16(obj.IdealClusterSize)
@@ -114,7 +118,7 @@ func (obj *Main) Init() error {
 	}
 
 	if obj.idealClusterSize < 1 {
-		return fmt.Errorf("IdealClusterSize should be at least one!")
+		return fmt.Errorf("the IdealClusterSize should be at least one")
 	}
 
 	if obj.NoServer && len(obj.Remotes) > 0 {
@@ -122,19 +126,19 @@ func (obj *Main) Init() error {
 		// here, so if we're okay with every remote graph running in an
 		// isolated mode, then this is okay. Improve on this if there's
 		// someone who really wants to be able to do this.
-		return fmt.Errorf("The Server is required when using Remotes!")
+		return fmt.Errorf("the Server is required when using Remotes")
 	}
 
 	if obj.CConns < 0 {
-		return fmt.Errorf("The CConns value should be at least zero!")
+		return fmt.Errorf("the CConns value should be at least zero")
 	}
 
 	if obj.ConvergedTimeout >= 0 && obj.CConns > 0 && len(obj.Remotes) > int(obj.CConns) {
-		return fmt.Errorf("You can't converge if you have more remotes than available connections!")
+		return fmt.Errorf("you can't converge if you have more remotes than available connections")
 	}
 
 	if obj.Depth < 0 { // user should not be using this argument manually
-		return fmt.Errorf("Negative values for Depth are not permitted!")
+		return fmt.Errorf("negative values for Depth are not permitted")
 	}
 
 	// transform the url list inputs into etcd typed lists
@@ -143,19 +147,19 @@ func (obj *Main) Init() error {
 		util.FlattenListWithSplit(obj.Seeds, []string{",", ";", " "}),
 	)
 	if err != nil && len(obj.Seeds) > 0 {
-		return fmt.Errorf("Seeds didn't parse correctly!")
+		return fmt.Errorf("the Seeds didn't parse correctly")
 	}
 	obj.clientURLs, err = etcdtypes.NewURLs(
 		util.FlattenListWithSplit(obj.ClientURLs, []string{",", ";", " "}),
 	)
 	if err != nil && len(obj.ClientURLs) > 0 {
-		return fmt.Errorf("ClientURLs didn't parse correctly!")
+		return fmt.Errorf("the ClientURLs didn't parse correctly")
 	}
 	obj.serverURLs, err = etcdtypes.NewURLs(
 		util.FlattenListWithSplit(obj.ServerURLs, []string{",", ";", " "}),
 	)
 	if err != nil && len(obj.ServerURLs) > 0 {
-		return fmt.Errorf("ServerURLs didn't parse correctly!")
+		return fmt.Errorf("the ServerURLs didn't parse correctly")
 	}
 
 	obj.exit = make(chan error)
@@ -195,10 +199,10 @@ func (obj *Main) Run() error {
 	if h := obj.Hostname; h != nil && *h != "" { // override by cli
 		hostname = *h
 	} else if err != nil {
-		return errwrap.Wrapf(err, "Can't get default hostname!")
+		return errwrap.Wrapf(err, "can't get default hostname")
 	}
 	if hostname == "" { // safety check
-		return fmt.Errorf("Hostname cannot be empty!")
+		return fmt.Errorf("hostname cannot be empty")
 	}
 
 	var prefix = fmt.Sprintf("/var/lib/%s/", obj.Program) // default prefix
@@ -210,24 +214,39 @@ func (obj *Main) Run() error {
 		if obj.TmpPrefix || obj.AllowTmpPrefix {
 			var err error
 			if prefix, err = ioutil.TempDir("", obj.Program+"-"+hostname+"-"); err != nil {
-				return fmt.Errorf("Main: Error: Can't create temporary prefix!")
+				return fmt.Errorf("can't create temporary prefix")
 			}
 			log.Println("Main: Warning: Working prefix directory is temporary!")
 
 		} else {
-			return fmt.Errorf("Main: Error: Can't create prefix!")
+			return fmt.Errorf("can't create prefix")
 		}
 	}
 	log.Printf("Main: Working prefix is: %s", prefix)
 	pgraphPrefix := fmt.Sprintf("%s/", path.Join(prefix, "pgraph")) // pgraph namespace
 	if err := os.MkdirAll(pgraphPrefix, 0770); err != nil {
-		return errwrap.Wrapf(err, "Can't create pgraph prefix")
+		return errwrap.Wrapf(err, "can't create pgraph prefix")
+	}
+
+	var prom *prometheus.Prometheus
+	if obj.Prometheus {
+		prom = &prometheus.Prometheus{
+			Listen: obj.PrometheusListen,
+		}
+		if err := prom.Init(); err != nil {
+			return errwrap.Wrapf(err, "can't create initiate Prometheus instance")
+		}
+
+		log.Printf("Main: Prometheus: Starting instance on %s", prom.Listen)
+		if err := prom.Start(); err != nil {
+			return errwrap.Wrapf(err, "can't start initiate Prometheus instance")
+		}
 	}
 
 	if !obj.NoPgp {
 		pgpPrefix := fmt.Sprintf("%s/", path.Join(prefix, "pgp"))
 		if err := os.MkdirAll(pgpPrefix, 0770); err != nil {
-			return errwrap.Wrapf(err, "Can't create pgp prefix")
+			return errwrap.Wrapf(err, "can't create pgp prefix")
 		}
 
 		pgpKeyringPath := path.Join(pgpPrefix, pgp.DefaultKeyringFile) // default path
@@ -238,7 +257,7 @@ func (obj *Main) Run() error {
 
 		var err error
 		if obj.pgpKeys, err = pgp.Import(pgpKeyringPath); err != nil && !os.IsNotExist(err) {
-			return errwrap.Wrapf(err, "Can't import pgp key")
+			return errwrap.Wrapf(err, "can't import pgp key")
 		}
 
 		if obj.pgpKeys == nil {
@@ -250,24 +269,23 @@ func (obj *Main) Run() error {
 
 			name, comment, email, err := pgp.ParseIdentity(identity)
 			if err != nil {
-				return errwrap.Wrapf(err, "Can't parse user string")
+				return errwrap.Wrapf(err, "can't parse user string")
 
 			}
 
 			// TODO: Make hash configurable
 			if obj.pgpKeys, err = pgp.Generate(name, comment, email, nil); err != nil {
-				return errwrap.Wrapf(err, "Can't creating pgp key")
+				return errwrap.Wrapf(err, "can't create pgp key")
 			}
 
 			if err := obj.pgpKeys.SaveKey(pgpKeyringPath); err != nil {
-				return errwrap.Wrapf(err, "Can't save pgp key")
+				return errwrap.Wrapf(err, "can't save pgp key")
 			}
 		}
 
 		// TODO: Import admin key
 	}
 
-	var wg sync.WaitGroup
 	var G, oldGraph *pgraph.Graph
 
 	// exit after `max-runtime` seconds for no reason at all...
@@ -308,7 +326,7 @@ func (obj *Main) Run() error {
 	)
 	if EmbdEtcd == nil {
 		// TODO: verify EmbdEtcd is not nil below...
-		obj.Exit(fmt.Errorf("Main: Etcd: Creation failed!"))
+		obj.Exit(fmt.Errorf("Main: Etcd: Creation failed"))
 	} else if err := EmbdEtcd.Startup(); err != nil { // startup (returns when etcd main loop is running)
 		obj.Exit(fmt.Errorf("Main: Etcd: Startup failed: %v", err))
 	}
@@ -325,43 +343,46 @@ func (obj *Main) Run() error {
 			return nil
 		}
 		// send our individual state into etcd for others to see
-		return etcd.EtcdSetHostnameConverged(EmbdEtcd, hostname, b) // TODO: what should happen on error?
+		return etcd.SetHostnameConverged(EmbdEtcd, hostname, b) // TODO: what should happen on error?
 	}
 	if EmbdEtcd != nil {
 		converger.SetStateFn(convergerStateFn)
 	}
 
+	// implementation of the World API (alternates can be substituted in)
+	world := &etcd.World{
+		Hostname: hostname,
+		EmbdEtcd: EmbdEtcd,
+	}
+
 	var gapiChan chan error // stream events are nil errors
 	if obj.GAPI != nil {
 		data := gapi.Data{
 			Hostname: hostname,
-			// NOTE: alternate implementations can be substituted in
-			World: &etcd.World{
-				Hostname: hostname,
-				EmbdEtcd: EmbdEtcd,
-			},
-			Noop:    obj.Noop,
-			NoWatch: obj.NoWatch,
+			World:    world,
+			Noop:     obj.Noop,
+			NoWatch:  obj.NoWatch,
 		}
 		if err := obj.GAPI.Init(data); err != nil {
 			obj.Exit(fmt.Errorf("Main: GAPI: Init failed: %v", err))
 		} else if !obj.NoWatch {
-			gapiChan = obj.GAPI.SwitchStream() // stream of graph switch events!
+			gapiChan = obj.GAPI.Next() // stream of graph switch events!
 		}
 	}
 
 	exitchan := make(chan struct{}) // exit on close
 	go func() {
 		startChan := make(chan struct{}) // start signal
-		go func() { startChan <- struct{}{} }()
+		close(startChan)                 // kick it off!
 
 		log.Println("Etcd: Starting...")
-		etcdChan := etcd.EtcdWatch(EmbdEtcd)
+		etcdChan := etcd.WatchAll(EmbdEtcd)
 		first := true // first loop or not
 		for {
 			log.Println("Main: Waiting...")
 			select {
 			case <-startChan: // kick the loop once at start
+				startChan = nil // disable
 				// pass
 
 			case b := <-etcdChan:
@@ -380,8 +401,7 @@ func (obj *Main) Run() error {
 				}
 				if err != nil {
 					obj.Exit(err) // trigger exit
-					continue
-					//return // TODO: return or wait for exitchan?
+					continue      // wait for exitchan
 				}
 				if obj.NoWatch { // extra safety for bad GAPI's
 					log.Printf("Main: GAPI stream should be quiet with NoWatch!") // fix the GAPI!
@@ -401,7 +421,7 @@ func (obj *Main) Run() error {
 			// run graph vertex LOCK...
 			if !first { // TODO: we can flatten this check out I think
 				converger.Pause() // FIXME: add sync wait?
-				G.Pause()         // sync
+				G.Pause(false)    // sync
 
 				//G.UnGroup() // FIXME: implement me if needed!
 			}
@@ -412,36 +432,46 @@ func (obj *Main) Run() error {
 				log.Printf("Config: Error creating new graph: %v", err)
 				// unpause!
 				if !first {
-					G.Start(&wg, first) // sync
-					converger.Start()   // after G.Start()
+					G.Start(first)    // sync
+					converger.Start() // after G.Start()
 				}
 				continue
 			}
 			newGraph.Flags = pgraph.Flags{Debug: obj.Flags.Debug}
 			// pass in the information we need
 			newGraph.AssociateData(&resources.Data{
-				Converger: converger,
-				Prefix:    pgraphPrefix,
-				Debug:     obj.Flags.Debug,
+				Hostname:   hostname,
+				Converger:  converger,
+				Prometheus: prom,
+				World:      world,
+				Prefix:     pgraphPrefix,
+				Debug:      obj.Flags.Debug,
 			})
 
-			// apply the global noop parameter if requested
-			if obj.Noop {
-				for _, m := range newGraph.GraphMetas() {
+			for _, m := range newGraph.GraphMetas() {
+				// apply the global noop parameter if requested
+				if obj.Noop {
 					m.Noop = obj.Noop
 				}
+
+				// append the semaphore to each resource
+				if obj.Sema > 0 { // NOTE: size == 0 would block
+					// a semaphore with an empty id is valid
+					m.Sema = append(m.Sema, fmt.Sprintf(":%d", obj.Sema))
+				}
 			}
 
 			// FIXME: make sure we "UnGroup()" any semi-destructive
 			// changes to the resources so our efficient GraphSync
 			// will be able to re-use and cmp to the old graph.
+			log.Printf("Main: GraphSync...")
 			newFullGraph, err := newGraph.GraphSync(oldGraph)
 			if err != nil {
 				log.Printf("Config: Error running graph sync: %v", err)
 				// unpause!
 				if !first {
-					G.Start(&wg, first) // sync
-					converger.Start()   // after G.Start()
+					G.Start(first)    // sync
+					converger.Start() // after G.Start()
 				}
 				continue
 			}
@@ -453,21 +483,31 @@ func (obj *Main) Run() error {
 			// TODO: do we want to do a transitive reduction?
 			// FIXME: run a type checker that verifies all the send->recv relationships
 
-			log.Printf("Graph: %v", G) // show graph
-			if obj.GraphvizFilter != "" {
-				if err := G.ExecGraphviz(obj.GraphvizFilter, obj.Graphviz); err != nil {
-					log.Printf("Graphviz: %v", err)
-				} else {
-					log.Printf("Graphviz: Successfully generated graph!")
-				}
+			// Call this here because at this point the graph does not
+			// know anything about the prometheus instance.
+			if err := prom.UpdatePgraphStartTime(); err != nil {
+				log.Printf("Main: Prometheus.UpdatePgraphStartTime() errored: %v", err)
 			}
 			// G.Start(...) needs to be synchronous or wait,
 			// because if half of the nodes are started and
 			// some are not ready yet and the EtcdWatch
 			// loops, we'll cause G.Pause(...) before we
 			// even got going, thus causing nil pointer errors
-			G.Start(&wg, first) // sync
-			converger.Start()   // after G.Start()
+			G.Start(first)    // sync
+			converger.Start() // after G.Start()
+
+			log.Printf("Graph: %v", G) // show graph
+			if obj.Graphviz != "" {
+				filter := obj.GraphvizFilter
+				if filter == "" {
+					filter = "dot" // directed graph default
+				}
+				if err := G.ExecGraphviz(filter, obj.Graphviz, hostname); err != nil {
+					log.Printf("Graphviz: %v", err)
+				} else {
+					log.Printf("Graphviz: Successfully generated graph!")
+				}
+			}
 			first = false
 		}
 	}()
@@ -492,7 +532,7 @@ func (obj *Main) Run() error {
 
 	// initialize the add watcher, which calls the f callback on map changes
 	convergerCb := func(f func(map[string]bool) error) (func(), error) {
-		return etcd.EtcdAddHostnameConvergedWatcher(EmbdEtcd, f)
+		return etcd.AddHostnameConvergedWatcher(EmbdEtcd, f)
 	}
 
 	// build remotes struct for remote ssh
@@ -531,34 +571,40 @@ func (obj *Main) Run() error {
 
 	if obj.GAPI != nil {
 		if err := obj.GAPI.Close(); err != nil {
-			err = errwrap.Wrapf(err, "GAPI closed poorly!")
+			err = errwrap.Wrapf(err, "the GAPI closed poorly")
 			reterr = multierr.Append(reterr, err) // list of errors
 		}
 	}
 
 	configWatcher.Close()                  // stop sending file changes to remotes
 	if err := remotes.Exit(); err != nil { // tell all the remote connections to shutdown; waits!
-		err = errwrap.Wrapf(err, "Remote exited poorly!")
+		err = errwrap.Wrapf(err, "the Remote exited poorly")
 		reterr = multierr.Append(reterr, err) // list of errors
 	}
 
-	G.Exit() // tell all the children to exit
-
 	// tell inner main loop to exit
 	close(exitchan)
 
+	G.Exit() // tells all the children to exit, and waits for them to do so
+
 	// cleanup etcd main loop last so it can process everything first
 	if err := EmbdEtcd.Destroy(); err != nil { // shutdown and cleanup etcd
-		err = errwrap.Wrapf(err, "Etcd exited poorly!")
+		err = errwrap.Wrapf(err, "embedded Etcd exited poorly")
 		reterr = multierr.Append(reterr, err) // list of errors
 	}
 
+	if obj.Prometheus {
+		log.Printf("Main: Prometheus: Stopping instance")
+		if err := prom.Stop(); err != nil {
+			err = errwrap.Wrapf(err, "the Prometheus instance exited poorly")
+			reterr = multierr.Append(reterr, err)
+		}
+	}
+
 	if obj.Flags.Debug {
 		log.Printf("Main: Graph: %v", G)
 	}
 
-	wg.Wait() // wait for primary go routines to exit
-
 	// TODO: wait for each vertex to exit...
 	log.Println("Goodbye!")
 	return reterr

main.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify

misc/delta-cpu.sh

@@ -0,0 +1,81 @@
+#!/bin/bash
+# shitty cpu count control, useful for live demos
+
+minimum=1	# don't decrease below this number of cpus
+maximum=8	# don't increase above this number of cpus
+count=1		# initial count
+factor=3
+function output() {
+count=$1	# arg!
+cat << EOF > ~/code/mgmt/examples/virt4.yaml
+---
+graph: mygraph
+resources:
+  virt:
+  - name: mgmt4
+    meta:
+      limit: .inf
+      burst: 0
+    uri: 'qemu:///session'
+    cpus: $count
+    maxcpus: $maximum
+    memory: 524288
+    boot:
+    - hd
+    disk:
+    - type: qcow2
+      source: "~/.local/share/libvirt/images/fedora-23-scratch.qcow2"
+    state: running
+    transient: false
+edges: []
+comment: "qemu-img create -b fedora-23.qcow2 -f qcow2 fedora-23-scratch.qcow2"
+EOF
+}
+#tput cuu 1 && tput el	# remove last line
+str=''
+tnuoc=$((maximum-count))	# backwards count
+count2=$((count * factor))
+tnuoc2=$((tnuoc * factor))
+left=`yes '>' | head -$count2 | paste -s -d '' -`
+right=`yes ' ' | head -$tnuoc2 | paste -s -d '' -`
+str="${left}${right}"
+_min=$((minimum-1))
+_max=$((maximum+1))
+reset	# clean up once...
+output $count	# call function
+while true; do
+
+	read -n1 -r -s -p "CPUs count is: $count; ${str} Press +/- key to adjust." key
+	if [ "$key" = "q" ] || [ "$key" = "Q" ]; then
+		echo	# newline
+		exit
+	fi
+	if [ ! "$key" = "+" ] && [ ! "$key" = "-" ] && [ ! "$key" = "=" ] && [ ! "$key" = "_" ]; then	# wrong key
+		reset	# woops, reset it all...
+		continue
+	fi
+	if [ "$key" == "+" ] || [ "$key" == "=" ]; then
+		count=$((count+1))
+	fi
+	if [ "$key" == "-" ] || [ "$key" == "_" ]; then
+		count=$((count-1))
+	fi
+	if [ $count -eq $_min ]; then	# min
+		count=$minimum
+	fi
+	if [ $count -eq $_max ]; then	# max
+		count=$maximum
+	fi
+
+	tnuoc=$((maximum-count))	# backwards count
+	#echo "count is: $count"
+	#echo "tnuoc is: $tnuoc"
+	count2=$((count * factor))
+	tnuoc2=$((tnuoc * factor))
+	left=`yes '>' | head -$count2 | paste -s -d '' -`
+	right=`yes ' ' | head -$tnuoc2 | paste -s -d '' -`
+	str="${left}${right}"
+	#echo "str is: $str"
+	echo -ne '\r'	# backup
+	output $count	# call function
+done

misc/make-deps.sh

@@ -14,26 +14,38 @@ sudo_command=$(which sudo)
 YUM=`which yum 2>/dev/null`
 DNF=`which dnf 2>/dev/null`
 APT=`which apt-get 2>/dev/null`
+BREW=`which brew 2>/dev/null`
+PACMAN=`which pacman 2>/dev/null`
 
 # if DNF is available use it
 if [ -x "$DNF" ]; then
 	YUM=$DNF
 fi
 
-if [ -z "$YUM" -a -z "$APT" ]; then
+if [ -z "$YUM" -a -z "$APT" -a -z "$BREW" -a -z "$PACMAN" ]; then
 	echo "The package managers can't be found."
 	exit 1
 fi
 
 if [ ! -z "$YUM" ]; then
 	$sudo_command $YUM install -y libvirt-devel
+	$sudo_command $YUM install -y augeas-devel
 
 fi
 if [ ! -z "$APT" ]; then
 	$sudo_command $APT install -y libvirt-dev || true
+	$sudo_command $APT install -y libaugeas-dev || true
 	$sudo_command $APT install -y libpcap0.8-dev || true
 fi
 
+if [ ! -z "$BREW" ]; then
+	$BREW install libvirt || true
+fi
+
+if [ ! -z "$PACMAN" ]; then
+	$sudo_command $PACMAN -S --noconfirm libvirt augeas libpcap
+fi
+
 if [ $travis -eq 0 ]; then
 	if [ ! -z "$YUM" ]; then
 		# some go dependencies are stored in mercurial
@@ -47,11 +59,14 @@ if [ $travis -eq 0 ]; then
 		$sudo_command $APT install -y golang-golang-x-tools || true
 		$sudo_command $APT install -y golang-go.tools || true
 	fi
+	if [ ! -z "$PACMAN" ]; then
+		$sudo_command $PACMAN -S --noconfirm go
+	fi
 fi
 
 # if golang is too old, we don't want to fail with an obscure error later
-if go version | grep 'go1\.[0123]\.'; then
-	echo "mgmt requires go1.4 or higher."
+if go version | grep 'go1\.[012345]\.'; then
+	echo "mgmt requires go1.6 or higher."
 	exit 1
 fi
 

misc/mgmt.service

@@ -5,7 +5,7 @@ After=systemd-networkd.service
 Requires=systemd-networkd.service
 
 [Service]
-ExecStart=/usr/bin/mgmt run ${OPTS}
+ExecStart=/usr/bin/mgmt run $OPTS
 RestartSec=5s
 Restart=always
 

pgp/pgp.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify

pgraph/actions.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -21,13 +21,17 @@ import (
 	"fmt"
 	"log"
 	"math"
+	"strings"
 	"sync"
 	"time"
 
 	"github.com/purpleidea/mgmt/event"
+	"github.com/purpleidea/mgmt/prometheus"
 	"github.com/purpleidea/mgmt/resources"
 
+	multierr "github.com/hashicorp/go-multierror"
 	errwrap "github.com/pkg/errors"
+	"golang.org/x/time/rate"
 )
 
 // GetTimestamp returns the timestamp of a vertex
@@ -60,29 +64,32 @@ func (g *Graph) OKTimestamp(v *Vertex) bool {
 	return true
 }
 
-// Poke notifies nodes after me in the dependency graph that they need refreshing...
-// NOTE: this assumes that this can never fail or need to be rescheduled
-func (g *Graph) Poke(v *Vertex, activity bool) error {
+// Poke tells nodes after me in the dependency graph that they need to refresh.
+func (g *Graph) Poke(v *Vertex) error {
+
+	// if we're pausing (or exiting) then we should suspend poke's so that
+	// the graph doesn't go on running forever until it's completely done!
+	// this is an optional feature which we can do by default on user exit
+	if g.fastPause {
+		return nil // TODO: should this be an error instead?
+	}
+
 	var wg sync.WaitGroup
 	// these are all the vertices pointing AWAY FROM v, eg: v -> ???
 	for _, n := range g.OutgoingGraphVertices(v) {
-		// XXX: if we're in state event and haven't been cancelled by
-		// apply, then we can cancel a poke to a child, right? XXX
-		// XXX: if n.Res.getState() != resources.ResStateEvent || activity { // is this correct?
-		if true || activity { // XXX: ???
+		// we can skip this poke if resource hasn't done work yet... it
+		// needs to be poked if already running, or not running though!
+		// TODO: does this need an || activity flag?
+		if n.Res.GetState() != resources.ResStateProcess {
 			if g.Flags.Debug {
 				log.Printf("%s[%s]: Poke: %s[%s]", v.Kind(), v.GetName(), n.Kind(), n.GetName())
 			}
 			wg.Add(1)
 			go func(nn *Vertex) error {
 				defer wg.Done()
-				edge := g.Adjacency[v][nn] // lookup
-				notify := edge.Notify && edge.Refresh()
-
-				// FIXME: is it okay that this is sync?
-				nn.SendEvent(event.EventPoke, true, notify)
-				// TODO: check return value?
-				return nil // never error for now...
+				//edge := g.Adjacency[v][nn] // lookup
+				//notify := edge.Notify && edge.Refresh()
+				return nn.SendEvent(event.EventPoke, nil)
 			}(n)
 
 		} else {
@@ -91,33 +98,41 @@ func (g *Graph) Poke(v *Vertex, activity bool) error {
 			}
 		}
 	}
+	// TODO: do something with return values?
 	wg.Wait() // wait for all the pokes to complete
 	return nil
 }
 
 // BackPoke pokes the pre-requisites that are stale and need to run before I can run.
 func (g *Graph) BackPoke(v *Vertex) {
+	var wg sync.WaitGroup
 	// these are all the vertices pointing TO v, eg: ??? -> v
 	for _, n := range g.IncomingGraphVertices(v) {
 		x, y, s := v.GetTimestamp(), n.GetTimestamp(), n.Res.GetState()
-		// if the parent timestamp needs poking AND it's not in state
-		// ResStateEvent, then poke it. If the parent is in ResStateEvent it
+		// If the parent timestamp needs poking AND it's not running
+		// Process, then poke it. If the parent is in ResStateProcess it
 		// means that an event is pending, so we'll be expecting a poke
 		// back soon, so we can safely discard the extra parent poke...
 		// TODO: implement a stateLT (less than) to tell if something
 		// happens earlier in the state cycle and that doesn't wrap nil
-		if x >= y && (s != resources.ResStateEvent && s != resources.ResStateCheckApply) {
+		if x >= y && (s != resources.ResStateProcess && s != resources.ResStateCheckApply) {
 			if g.Flags.Debug {
 				log.Printf("%s[%s]: BackPoke: %s[%s]", v.Kind(), v.GetName(), n.Kind(), n.GetName())
 			}
-			// FIXME: is it okay that this is sync?
-			n.SendEvent(event.EventBackPoke, true, false)
+			wg.Add(1)
+			go func(nn *Vertex) error {
+				defer wg.Done()
+				return nn.SendEvent(event.EventBackPoke, nil)
+			}(n)
+
 		} else {
 			if g.Flags.Debug {
 				log.Printf("%s[%s]: BackPoke: %s[%s]: Skipped!", v.Kind(), v.GetName(), n.Kind(), n.GetName())
 			}
 		}
 	}
+	// TODO: do something with return values?
+	wg.Wait() // wait for all the pokes to complete
 }
 
 // RefreshPending determines if any previous nodes have a refresh pending here.
@@ -159,114 +174,156 @@ func (g *Graph) Process(v *Vertex) error {
 	if g.Flags.Debug {
 		log.Printf("%s[%s]: Process()", obj.Kind(), obj.GetName())
 	}
-	obj.SetState(resources.ResStateEvent)
-	var ok = true
-	var applied = false // did we run an apply?
+	// FIXME: should these SetState methods be here or after the sema code?
+	defer obj.SetState(resources.ResStateNil) // reset state when finished
+	obj.SetState(resources.ResStateProcess)
+
 	// is it okay to run dependency wise right now?
 	// if not, that's okay because when the dependency runs, it will poke
 	// us back and we will run if needed then!
-	if g.OKTimestamp(v) {
-		if g.Flags.Debug {
-			log.Printf("%s[%s]: OKTimestamp(%v)", obj.Kind(), obj.GetName(), v.GetTimestamp())
+	if !g.OKTimestamp(v) {
+		go g.BackPoke(v)
+		return nil
+	}
+	// timestamp must be okay...
+	if g.Flags.Debug {
+		log.Printf("%s[%s]: OKTimestamp(%v)", obj.Kind(), obj.GetName(), v.GetTimestamp())
+	}
+
+	// semaphores!
+	// These shouldn't ever block an exit, since the graph should eventually
+	// converge causing their them to unlock. More interestingly, since they
+	// run in a DAG alphabetically, there is no way to permanently deadlock,
+	// assuming that resources individually don't ever block from finishing!
+	// The exception is that semaphores with a zero count will always block!
+	// TODO: Add a close mechanism to close/unblock zero count semaphores...
+	semas := obj.Meta().Sema
+	if g.Flags.Debug && len(semas) > 0 {
+		log.Printf("%s[%s]: Sema: P(%s)", obj.Kind(), obj.GetName(), strings.Join(semas, ", "))
+	}
+	if err := g.SemaLock(semas); err != nil { // lock
+		// NOTE: in practice, this might not ever be truly necessary...
+		return fmt.Errorf("shutdown of semaphores")
+	}
+	defer g.SemaUnlock(semas) // unlock
+	if g.Flags.Debug && len(semas) > 0 {
+		defer log.Printf("%s[%s]: Sema: V(%s)", obj.Kind(), obj.GetName(), strings.Join(semas, ", "))
+	}
+
+	var ok = true
+	var applied = false // did we run an apply?
+
+	// connect any senders to receivers and detect if values changed
+	if updated, err := obj.SendRecv(obj); err != nil {
+		return errwrap.Wrapf(err, "could not SendRecv in Process")
+	} else if len(updated) > 0 {
+		for _, changed := range updated {
+			if changed { // at least one was updated
+				obj.StateOK(false) // invalidate cache, mark as dirty
+				break
+			}
 		}
+	}
 
-		obj.SetState(resources.ResStateCheckApply)
+	var noop = obj.Meta().Noop // lookup the noop value
+	var refresh bool
+	var checkOK bool
+	var err error
 
-		// connect any senders to receivers and detect if values changed
-		if updated, err := obj.SendRecv(obj); err != nil {
-			return errwrap.Wrapf(err, "could not SendRecv in Process")
-		} else if len(updated) > 0 {
-			for _, changed := range updated {
-				if changed { // at least one was updated
-					obj.StateOK(false) // invalidate cache, mark as dirty
-					break
+	if g.Flags.Debug {
+		log.Printf("%s[%s]: CheckApply(%t)", obj.Kind(), obj.GetName(), !noop)
+	}
+
+	// lookup the refresh (notification) variable
+	refresh = g.RefreshPending(v) // do i need to perform a refresh?
+	obj.SetRefresh(refresh)       // tell the resource
+
+	// changes can occur after this...
+	obj.SetState(resources.ResStateCheckApply)
+
+	// check cached state, to skip CheckApply; can't skip if refreshing
+	if !refresh && obj.IsStateOK() {
+		checkOK, err = true, nil
+
+		// NOTE: technically this block is wrong because we don't know
+		// if the resource implements refresh! If it doesn't, we could
+		// skip this, but it doesn't make a big difference under noop!
+	} else if noop && refresh { // had a refresh to do w/ noop!
+		checkOK, err = false, nil // therefore the state is wrong
+
+		// run the CheckApply!
+	} else {
+		// if this fails, don't UpdateTimestamp()
+		checkOK, err = obj.CheckApply(!noop)
+
+		if promErr := obj.Prometheus().UpdateCheckApplyTotal(obj.Kind(), !noop, !checkOK, err != nil); promErr != nil {
+			// TODO: how to error correctly
+			log.Printf("%s[%s]: Prometheus.UpdateCheckApplyTotal() errored: %v", v.Kind(), v.GetName(), err)
+		}
+		// TODO: Can the `Poll` converged timeout tracking be a
+		// more general method for all converged timeouts? this
+		// would simplify the resources by removing boilerplate
+		if v.Meta().Poll > 0 {
+			if !checkOK { // something changed, restart timer
+				cuid, _, _ := v.Res.ConvergerUIDs() // get the converger uid used to report status
+				cuid.ResetTimer()                   // activity!
+				if g.Flags.Debug {
+					log.Printf("%s[%s]: Converger: ResetTimer", obj.Kind(), obj.GetName())
 				}
 			}
 		}
-
-		var noop = obj.Meta().Noop // lookup the noop value
-		var refresh bool
-		var checkOK bool
-		var err error
-
-		if g.Flags.Debug {
-			log.Printf("%s[%s]: CheckApply(%t)", obj.Kind(), obj.GetName(), !noop)
-		}
-
-		// lookup the refresh (notification) variable
-		refresh = g.RefreshPending(v) // do i need to perform a refresh?
-		obj.SetRefresh(refresh)       // tell the resource
-
-		// check cached state, to skip CheckApply; can't skip if refreshing
-		if !refresh && obj.IsStateOK() {
-			checkOK, err = true, nil
-
-			// NOTE: technically this block is wrong because we don't know
-			// if the resource implements refresh! If it doesn't, we could
-			// skip this, but it doesn't make a big difference under noop!
-		} else if noop && refresh { // had a refresh to do w/ noop!
-			checkOK, err = false, nil // therefore the state is wrong
-
-			// run the CheckApply!
-		} else {
-			// if this fails, don't UpdateTimestamp()
-			checkOK, err = obj.CheckApply(!noop)
-		}
-
-		if checkOK && err != nil { // should never return this way
-			log.Fatalf("%s[%s]: CheckApply(): %t, %+v", obj.Kind(), obj.GetName(), checkOK, err)
-		}
-		if g.Flags.Debug {
-			log.Printf("%s[%s]: CheckApply(): %t, %v", obj.Kind(), obj.GetName(), checkOK, err)
-		}
-
-		// if CheckApply ran without noop and without error, state should be good
-		if !noop && err == nil { // aka !noop || checkOK
-			obj.StateOK(true) // reset
-			if refresh {
-				g.SetUpstreamRefresh(v, false) // refresh happened, clear the request
-				obj.SetRefresh(false)
-			}
-		}
-
-		if !checkOK { // if state *was* not ok, we had to have apply'ed
-			if err != nil { // error during check or apply
-				ok = false
-			} else {
-				applied = true
-			}
-		}
-
-		// when noop is true we always want to update timestamp
-		if noop && err == nil {
-			ok = true
-		}
-
-		if ok {
-			// did we actually do work?
-			activity := applied
-			if noop {
-				activity = false // no we didn't do work...
-			}
-
-			if activity { // add refresh flag to downstream edges...
-				g.SetDownstreamRefresh(v, true)
-			}
-
-			// update this timestamp *before* we poke or the poked
-			// nodes might fail due to having a too old timestamp!
-			v.UpdateTimestamp()                    // this was touched...
-			obj.SetState(resources.ResStatePoking) // can't cancel parent poke
-			if err := g.Poke(v, activity); err != nil {
-				return errwrap.Wrapf(err, "the Poke() failed")
-			}
-		}
-		// poke at our pre-req's instead since they need to refresh/run...
-		return errwrap.Wrapf(err, "could not Process() successfully")
 	}
-	// else... only poke at the pre-req's that need to run
-	go g.BackPoke(v)
-	return nil
+
+	if checkOK && err != nil { // should never return this way
+		log.Fatalf("%s[%s]: CheckApply(): %t, %+v", obj.Kind(), obj.GetName(), checkOK, err)
+	}
+	if g.Flags.Debug {
+		log.Printf("%s[%s]: CheckApply(): %t, %v", obj.Kind(), obj.GetName(), checkOK, err)
+	}
+
+	// if CheckApply ran without noop and without error, state should be good
+	if !noop && err == nil { // aka !noop || checkOK
+		obj.StateOK(true) // reset
+		if refresh {
+			g.SetUpstreamRefresh(v, false) // refresh happened, clear the request
+			obj.SetRefresh(false)
+		}
+	}
+
+	if !checkOK { // if state *was* not ok, we had to have apply'ed
+		if err != nil { // error during check or apply
+			ok = false
+		} else {
+			applied = true
+		}
+	}
+
+	// when noop is true we always want to update timestamp
+	if noop && err == nil {
+		ok = true
+	}
+
+	if ok {
+		// did we actually do work?
+		activity := applied
+		if noop {
+			activity = false // no we didn't do work...
+		}
+
+		if activity { // add refresh flag to downstream edges...
+			g.SetDownstreamRefresh(v, true)
+		}
+
+		// update this timestamp *before* we poke or the poked
+		// nodes might fail due to having a too old timestamp!
+		v.UpdateTimestamp()                    // this was touched...
+		obj.SetState(resources.ResStatePoking) // can't cancel parent poke
+		if err := g.Poke(v); err != nil {
+			return errwrap.Wrapf(err, "the Poke() failed")
+		}
+	}
+	// poke at our pre-req's instead since they need to refresh/run...
+	return errwrap.Wrapf(err, "could not Process() successfully")
 }
 
 // SentinelErr is a sentinal as an error type that wraps an arbitrary error.
@@ -279,6 +336,180 @@ func (obj *SentinelErr) Error() string {
 	return obj.err.Error()
 }
 
+// innerWorker is the CheckApply runner that reads from processChan.
+// TODO: would it be better if this was a method on BaseRes that took in *Graph?
+func (g *Graph) innerWorker(v *Vertex) {
+	obj := v.Res
+	running := false
+	done := make(chan struct{})
+	playback := false                      // do we need to run another one?
+	_, wcuid, pcuid := obj.ConvergerUIDs() // get extra cuids (worker, process)
+
+	waiting := false
+	var timer = time.NewTimer(time.Duration(math.MaxInt64)) // longest duration
+	if !timer.Stop() {
+		<-timer.C // unnecessary, shouldn't happen
+	}
+
+	var delay = time.Duration(v.Meta().Delay) * time.Millisecond
+	var retry = v.Meta().Retry // number of tries left, -1 for infinite
+	var limiter = rate.NewLimiter(v.Meta().Limit, v.Meta().Burst)
+	limited := false
+
+	wg := &sync.WaitGroup{} // wait for Process routine to exit
+
+Loop:
+	for {
+		select {
+		case ev, ok := <-obj.ProcessChan(): // must use like this
+			if !ok { // processChan closed, let's exit
+				break Loop // no event, so no ack!
+			}
+			if v.Res.Meta().Poll == 0 { // skip for polling
+				wcuid.SetConverged(false)
+			}
+
+			// if process started, but no action yet, skip!
+			if v.Res.GetState() == resources.ResStateProcess {
+				if g.Flags.Debug {
+					log.Printf("%s[%s]: Skipped event!", v.Kind(), v.GetName())
+				}
+				ev.ACK() // ready for next message
+				v.Res.QuiesceGroup().Done()
+				continue
+			}
+
+			// if running, we skip running a new execution!
+			// if waiting, we skip running a new execution!
+			if running || waiting {
+				if g.Flags.Debug {
+					log.Printf("%s[%s]: Playback added!", v.Kind(), v.GetName())
+				}
+				playback = true
+				ev.ACK() // ready for next message
+				v.Res.QuiesceGroup().Done()
+				continue
+			}
+
+			// catch invalid rates
+			if v.Meta().Burst == 0 && !(v.Meta().Limit == rate.Inf) { // blocked
+				e := fmt.Errorf("%s[%s]: Permanently limited (rate != Inf, burst: 0)", v.Kind(), v.GetName())
+				ev.ACK() // ready for next message
+				v.Res.QuiesceGroup().Done()
+				v.SendEvent(event.EventExit, &SentinelErr{e})
+				continue
+			}
+
+			// rate limit
+			// FIXME: consider skipping rate limit check if
+			// the event is a poke instead of a watch event
+			if !limited && !(v.Meta().Limit == rate.Inf) { // skip over the playback event...
+				now := time.Now()
+				r := limiter.ReserveN(now, 1) // one event
+				// r.OK() seems to always be true here!
+				d := r.DelayFrom(now)
+				if d > 0 { // delay
+					limited = true
+					playback = true
+					log.Printf("%s[%s]: Limited (rate: %v/sec, burst: %d, next: %v)", v.Kind(), v.GetName(), v.Meta().Limit, v.Meta().Burst, d)
+					// start the timer...
+					timer.Reset(d)
+					waiting = true // waiting for retry timer
+					ev.ACK()
+					v.Res.QuiesceGroup().Done()
+					continue
+				} // otherwise, we run directly!
+			}
+			limited = false // let one through
+
+			wg.Add(1)
+			running = true
+			go func(ev *event.Event) {
+				pcuid.SetConverged(false) // "block" Process
+				defer wg.Done()
+				if e := g.Process(v); e != nil {
+					playback = true
+					log.Printf("%s[%s]: CheckApply errored: %v", v.Kind(), v.GetName(), e)
+					if retry == 0 {
+						if err := obj.Prometheus().UpdateState(fmt.Sprintf("%v[%v]", v.Kind(), v.GetName()), v.Kind(), prometheus.ResStateHardFail); err != nil {
+							// TODO: how to error this?
+							log.Printf("%s[%s]: Prometheus.UpdateState() errored: %v", v.Kind(), v.GetName(), err)
+						}
+
+						// wrap the error in the sentinel
+						v.Res.QuiesceGroup().Done() // before the Wait that happens in SendEvent!
+						v.SendEvent(event.EventExit, &SentinelErr{e})
+						return
+					}
+					if retry > 0 { // don't decrement the -1
+						retry--
+					}
+					if err := obj.Prometheus().UpdateState(fmt.Sprintf("%v[%v]", v.Kind(), v.GetName()), v.Kind(), prometheus.ResStateSoftFail); err != nil {
+						// TODO: how to error this?
+						log.Printf("%s[%s]: Prometheus.UpdateState() errored: %v", v.Kind(), v.GetName(), err)
+					}
+					log.Printf("%s[%s]: CheckApply: Retrying after %.4f seconds (%d left)", v.Kind(), v.GetName(), delay.Seconds(), retry)
+					// start the timer...
+					timer.Reset(delay)
+					waiting = true // waiting for retry timer
+					// don't v.Res.QuiesceGroup().Done() b/c
+					// the timer is running and it can exit!
+					return
+				}
+				retry = v.Meta().Retry // reset on success
+				close(done)            // trigger
+			}(ev)
+			ev.ACK() // sync (now mostly useless)
+
+		case <-timer.C:
+			if v.Res.Meta().Poll == 0 { // skip for polling
+				wcuid.SetConverged(false)
+			}
+			waiting = false
+			if !timer.Stop() {
+				//<-timer.C // blocks, docs are wrong!
+			}
+			log.Printf("%s[%s]: CheckApply delay expired!", v.Kind(), v.GetName())
+			close(done)
+
+		// a CheckApply run (with possibly retry pause) finished
+		case <-done:
+			if v.Res.Meta().Poll == 0 { // skip for polling
+				wcuid.SetConverged(false)
+			}
+			if g.Flags.Debug {
+				log.Printf("%s[%s]: CheckApply finished!", v.Kind(), v.GetName())
+			}
+			done = make(chan struct{}) // reset
+			// re-send this event, to trigger a CheckApply()
+			if playback {
+				// this lock avoids us sending to
+				// channel after we've closed it!
+				// TODO: can this experience indefinite postponement ?
+				// see: https://github.com/golang/go/issues/11506
+				// pause or exit is in process if not quiescing!
+				if !v.Res.IsQuiescing() {
+					playback = false
+					v.Res.QuiesceGroup().Add(1) // lock around it, b/c still running...
+					go func() {
+						obj.Event() // replay a new event
+						v.Res.QuiesceGroup().Done()
+					}()
+				}
+			}
+			running = false
+			pcuid.SetConverged(true) // "unblock" Process
+			v.Res.QuiesceGroup().Done()
+
+		case <-wcuid.ConvergedTimer():
+			wcuid.SetConverged(true) // converged!
+			continue
+		}
+	}
+	wg.Wait()
+	return
+}
+
 // Worker is the common run frontend of the vertex. It handles all of the retry
 // and retry delay common code, and ultimately returns the final status of this
 // vertex execution.
@@ -289,68 +520,36 @@ func (g *Graph) Worker(v *Vertex) error {
 	// the Watch() function about which graph it is
 	// running on, which isolates things nicely...
 	obj := v.Res
-	processChan := make(chan event.Event)
+	if g.Flags.Debug {
+		log.Printf("%s[%s]: Worker: Running", v.Kind(), v.GetName())
+		defer log.Printf("%s[%s]: Worker: Stopped", v.Kind(), v.GetName())
+	}
+	// run the init (should match 1-1 with Close function)
+	if err := obj.Init(); err != nil {
+		obj.ProcessExit()
+		// always exit the worker function by finishing with Close()
+		if e := obj.Close(); e != nil {
+			err = multierr.Append(err, e) // list of errors
+		}
+		return errwrap.Wrapf(err, "could not Init() resource")
+	}
+
+	// if the CheckApply run takes longer than the converged
+	// timeout, we could inappropriately converge mid-apply!
+	// avoid this by blocking convergence with a fake report
+	// we also add a similar blocker around the worker loop!
+	_, wcuid, pcuid := obj.ConvergerUIDs() // get extra cuids (worker, process)
+	// XXX: put these in Init() ?
+	wcuid.SetConverged(true) // starts off false, and waits for loop timeout
+	pcuid.SetConverged(true) // starts off true, because it's not running...
+
+	wg := obj.ProcessSync()
+	wg.Add(1)
 	go func() {
-		running := false
-		var timer = time.NewTimer(time.Duration(math.MaxInt64)) // longest duration
-		if !timer.Stop() {
-			<-timer.C // unnecessary, shouldn't happen
-		}
-		var delay = time.Duration(v.Meta().Delay) * time.Millisecond
-		var retry = v.Meta().Retry // number of tries left, -1 for infinite
-		var saved event.Event
-	Loop:
-		for {
-			// this has to be synchronous, because otherwise the Res
-			// event loop will keep running and change state,
-			// causing the converged timeout to fire!
-			select {
-			case event, ok := <-processChan: // must use like this
-				if running && ok {
-					// we got an event that wasn't a close,
-					// while we were waiting for the timer!
-					// if this happens, it might be a bug:(
-					log.Fatalf("%s[%s]: Worker: Unexpected event: %+v", v.Kind(), v.GetName(), event)
-				}
-				if !ok { // processChan closed, let's exit
-					break Loop // no event, so no ack!
-				}
-
-				// the above mentioned synchronous part, is the
-				// running of this function, paired with an ack.
-				if e := g.Process(v); e != nil {
-					saved = event
-					log.Printf("%s[%s]: CheckApply errored: %v", v.Kind(), v.GetName(), e)
-					if retry == 0 {
-						// wrap the error in the sentinel
-						event.ACKNACK(&SentinelErr{e}) // fail the Watch()
-						break Loop
-					}
-					if retry > 0 { // don't decrement the -1
-						retry--
-					}
-					log.Printf("%s[%s]: CheckApply: Retrying after %.4f seconds (%d left)", v.Kind(), v.GetName(), delay.Seconds(), retry)
-					// start the timer...
-					timer.Reset(delay)
-					running = true
-					continue
-				}
-				retry = v.Meta().Retry // reset on success
-				event.ACK()            // sync
-
-			case <-timer.C:
-				if !timer.Stop() {
-					//<-timer.C // blocks, docs are wrong!
-				}
-				running = false
-				log.Printf("%s[%s]: CheckApply delay expired!", v.Kind(), v.GetName())
-				// re-send this failed event, to trigger a CheckApply()
-				go func() { processChan <- saved }()
-				// TODO: should we send a fake event instead?
-				//saved = nil
-			}
-		}
+		defer wg.Done()
+		g.innerWorker(v)
 	}()
+
 	var err error // propagate the error up (this is a permanent BAD error!)
 	// the watch delay runs inside of the Watch resource loop, so that it
 	// can still process signals and exit if needed. It shouldn't run any
@@ -378,8 +577,15 @@ func (g *Graph) Worker(v *Vertex) error {
 				case event := <-obj.Events():
 					// NOTE: this code should match the similar Res code!
 					//cuid.SetConverged(false) // TODO: ?
-					if exit, send := obj.ReadEvent(&event); exit {
-						return nil // exit
+					if exit, send := obj.ReadEvent(event); exit != nil {
+						obj.ProcessExit()
+						err := *exit // exit err
+						if e := obj.Close(); err == nil {
+							err = e
+						} else if e != nil {
+							err = multierr.Append(err, e) // list of errors
+						}
+						return err // exit
 					} else if send {
 						// if we dive down this rabbit hole, our
 						// timer.C won't get seen until we get out!
@@ -408,14 +614,22 @@ func (g *Graph) Worker(v *Vertex) error {
 			// NOTE: we can avoid the send if running Watch guarantees
 			// one CheckApply event on startup!
 			//if pendingSendEvent { // TODO: should this become a list in the future?
-			//	if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
+			//	if err := obj.Event() err != nil {
 			//		return err // we exit or bubble up a NACK...
 			//	}
 			//}
 		}
 
 		// TODO: reset the watch retry count after some amount of success
-		e := v.Res.Watch(processChan)
+		var e error
+		if v.Res.Meta().Poll > 0 { // poll instead of watching :(
+			cuid, _, _ := v.Res.ConvergerUIDs() // get the converger uid used to report status
+			cuid.StartTimer()
+			e = v.Res.Poll()
+			cuid.StopTimer() // clean up nicely
+		} else {
+			e = v.Res.Watch() // run the watch normally
+		}
 		if e == nil { // exit signal
 			err = nil // clean exit
 			break
@@ -439,37 +653,40 @@ func (g *Graph) Worker(v *Vertex) error {
 		// by getting the Watch resource to send one event once it's up!
 		//v.SendEvent(eventPoke, false, false)
 	}
-	close(processChan)
+
+	obj.ProcessExit()
+	// close resource and return possible errors if any
+	if e := obj.Close(); err == nil {
+		err = e
+	} else if e != nil {
+		err = multierr.Append(err, e) // list of errors
+	}
 	return err
 }
 
 // Start is a main kick to start the graph. It goes through in reverse topological
 // sort order so that events can't hit un-started vertices.
-func (g *Graph) Start(wg *sync.WaitGroup, first bool) { // start or continue
+func (g *Graph) Start(first bool) { // start or continue
 	log.Printf("State: %v -> %v", g.setState(graphStateStarting), g.getState())
 	defer log.Printf("State: %v -> %v", g.setState(graphStateStarted), g.getState())
 	t, _ := g.TopologicalSort()
-	// TODO: only calculate indegree if `first` is true to save resources
 	indegree := g.InDegree() // compute all of the indegree's
-	for _, v := range Reverse(t) {
-
-		if !v.Res.IsWatching() { // if Watch() is not running...
-			wg.Add(1)
-			// must pass in value to avoid races...
-			// see: https://ttboj.wordpress.com/2015/07/27/golang-parallelism-issues-causing-too-many-open-files-error/
-			go func(vv *Vertex) {
-				defer wg.Done()
-				// TODO: if a sufficient number of workers error,
-				// should something be done? Will these restart
-				// after perma-failure if we have a graph change?
-				if err := g.Worker(vv); err != nil { // contains the Watch and CheckApply loops
-					log.Printf("%s[%s]: Exited with failure: %v", vv.Kind(), vv.GetName(), err)
-					return
-				}
-				log.Printf("%s[%s]: Exited", vv.Kind(), vv.GetName())
-			}(v)
-		}
+	reversed := Reverse(t)
+	wg := &sync.WaitGroup{}
+	for _, v := range reversed { // run the Setup() for everyone first
+		// run these in parallel, as long as we wait before continuing
+		wg.Add(1)
+		go func(vv *Vertex) {
+			defer wg.Done()
+			if !vv.Res.IsWorking() { // if Worker() is not running...
+				vv.Res.Setup() // initialize some vars in the resource
+			}
+		}(v)
+	}
+	wg.Wait()
 
+	// run through the topological reverse, and start or unpause each vertex
+	for _, v := range reversed {
 		// selective poke: here we reduce the number of initial pokes
 		// to the minimum required to activate every vertex in the
 		// graph, either by direct action, or by getting poked by a
@@ -482,36 +699,75 @@ func (g *Graph) Start(wg *sync.WaitGroup, first bool) { // start or continue
 		// if we are unpausing (since it's not the first run of this
 		// function) we need to poke to *unpause* every graph vertex,
 		// and not just selectively the subset with no indegree.
-		if (!first) || indegree[v] == 0 {
-			// ensure state is started before continuing on to next vertex
-			for !v.SendEvent(event.EventStart, true, false) {
-				if g.Flags.Debug {
-					// if SendEvent fails, we aren't up yet
-					log.Printf("%s[%s]: Retrying SendEvent(Start)", v.Kind(), v.GetName())
-					// sleep here briefly or otherwise cause
-					// a different goroutine to be scheduled
-					time.Sleep(1 * time.Millisecond)
+
+		// let the startup code know to poke or not
+		// this triggers a CheckApply AFTER Watch is Running()
+		// We *don't* need to also do this to new nodes or nodes that
+		// are about to get unpaused, because they'll get poked by one
+		// of the indegree == 0 vertices, and an important aspect of the
+		// Process() function is that even if the state is correct, it
+		// will pass through the Poke so that it flows through the DAG.
+		v.Res.Starter(indegree[v] == 0)
+
+		var unpause = true
+		if !v.Res.IsWorking() { // if Worker() is not running...
+			unpause = false // doesn't need unpausing on first start
+			g.wg.Add(1)
+			// must pass in value to avoid races...
+			// see: https://ttboj.wordpress.com/2015/07/27/golang-parallelism-issues-causing-too-many-open-files-error/
+			go func(vv *Vertex) {
+				defer g.wg.Done()
+				defer v.Res.Reset()
+				// TODO: if a sufficient number of workers error,
+				// should something be done? Should these restart
+				// after perma-failure if we have a graph change?
+				log.Printf("%s[%s]: Started", vv.Kind(), vv.GetName())
+				if err := g.Worker(vv); err != nil { // contains the Watch and CheckApply loops
+					log.Printf("%s[%s]: Exited with failure: %v", vv.Kind(), vv.GetName(), err)
+					return
 				}
-			}
+				log.Printf("%s[%s]: Exited", vv.Kind(), vv.GetName())
+			}(v)
+		}
+
+		select {
+		case <-v.Res.Started(): // block until started
+		case <-v.Res.Stopped(): // we failed on init
+			// if the resource Init() fails, we don't hang!
+		}
+
+		if unpause { // unpause (if needed)
+			v.Res.SendEvent(event.EventStart, nil) // sync!
 		}
 	}
+	// we wait for everyone to start before exiting!
 }
 
-// Pause sends pause events to the graph in a topological sort order.
-func (g *Graph) Pause() {
+// Pause sends pause events to the graph in a topological sort order. If you set
+// the fastPause argument to true, then it will ask future propagation waves to
+// not run through the graph before exiting, and instead will exit much quicker.
+func (g *Graph) Pause(fastPause bool) {
 	log.Printf("State: %v -> %v", g.setState(graphStatePausing), g.getState())
 	defer log.Printf("State: %v -> %v", g.setState(graphStatePaused), g.getState())
+	if fastPause {
+		g.fastPause = true // set flag
+	}
 	t, _ := g.TopologicalSort()
 	for _, v := range t { // squeeze out the events...
-		v.SendEvent(event.EventPause, true, false)
+		v.SendEvent(event.EventPause, nil) // sync
 	}
+	g.fastPause = false // reset flag
 }
 
 // Exit sends exit events to the graph in a topological sort order.
 func (g *Graph) Exit() {
-	if g == nil {
+	if g == nil { // empty graph that wasn't populated yet
 		return
-	} // empty graph that wasn't populated yet
+	}
+
+	// FIXME: a second ^C could put this into fast pause, but do it for now!
+	g.Pause(true) // implement this with pause to avoid duplicating the code
+
 	t, _ := g.TopologicalSort()
 	for _, v := range t { // squeeze out the events...
 		// turn off the taps...
@@ -520,6 +776,8 @@ func (g *Graph) Exit() {
 		// when we hit the 'default' in the select statement!
 		// XXX: we can do this to quiesce, but it's not necessary now
 
-		v.SendEvent(event.EventExit, true, false)
+		v.SendEvent(event.EventExit, nil)
+		v.Res.WaitGroup().Wait()
 	}
+	g.wg.Wait() // for now, this doesn't need to be a separate Wait() method
 }

pgraph/autoedge.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -45,7 +45,7 @@ func (g *Graph) addEdgesByMatchingUIDS(v *Vertex, uids []resources.ResUID) []boo
 			// that is to say, the name value of a res is a helpful
 			// handle, but it is not necessarily a unique identity!
 			// remember, resources can return multiple UID's each!
-			if resources.UIDExistsInUIDs(uid, vv.GetUIDs()) {
+			if resources.UIDExistsInUIDs(uid, vv.UIDs()) {
 				// add edge from: vv -> v
 				if uid.Reversed() {
 					txt := fmt.Sprintf("AutoEdge: %v[%v] -> %v[%v]", vv.Kind(), vv.GetName(), v.Kind(), v.GetName())

pgraph/autogroup.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -54,7 +54,7 @@ func (ag *baseGrouper) name() string {
 // the name method is the only exception: call it any time without side effects!
 func (ag *baseGrouper) init(g *Graph) error {
 	if ag.graph != nil {
-		return fmt.Errorf("The init method has already been called!")
+		return fmt.Errorf("the init method has already been called")
 	}
 	ag.graph = g                               // pointer
 	ag.vertices = ag.graph.GetVerticesSorted() // cache in deterministic order!
@@ -108,27 +108,27 @@ func (ag *baseGrouper) vertexNext() (v1, v2 *Vertex, err error) {
 
 func (ag *baseGrouper) vertexCmp(v1, v2 *Vertex) error {
 	if v1 == nil || v2 == nil {
-		return fmt.Errorf("Vertex is nil!")
+		return fmt.Errorf("the vertex is nil")
 	}
 	if v1 == v2 { // skip yourself
-		return fmt.Errorf("Vertices are the same!")
+		return fmt.Errorf("the vertices are the same")
 	}
 	if v1.Kind() != v2.Kind() { // we must group similar kinds
 		// TODO: maybe future resources won't need this limitation?
-		return fmt.Errorf("The two resources aren't the same kind!")
+		return fmt.Errorf("the two resources aren't the same kind")
 	}
 	// someone doesn't want to group!
 	if !v1.Meta().AutoGroup || !v2.Meta().AutoGroup {
-		return fmt.Errorf("One of the autogroup flags is false!")
+		return fmt.Errorf("one of the autogroup flags is false")
 	}
 	if v1.Res.IsGrouped() { // already grouped!
-		return fmt.Errorf("Already grouped!")
+		return fmt.Errorf("already grouped")
 	}
 	if len(v2.Res.GetGroup()) > 0 { // already has children grouped!
-		return fmt.Errorf("Already has groups!")
+		return fmt.Errorf("already has groups")
 	}
 	if !v1.Res.GroupCmp(v2.Res) { // resource groupcmp failed!
-		return fmt.Errorf("The GroupCmp failed!")
+		return fmt.Errorf("the GroupCmp failed")
 	}
 	return nil // success
 }
@@ -173,7 +173,7 @@ func (ag *nonReachabilityGrouper) vertexNext() (v1, v2 *Vertex, err error) {
 	for {
 		v1, v2, err = ag.baseGrouper.vertexNext() // get all iterable pairs
 		if err != nil {
-			log.Fatalf("Error running autoGroup(vertexNext): %v", err)
+			log.Fatalf("error running autoGroup(vertexNext): %v", err)
 		}
 
 		if v1 != v2 { // ignore self cmp early (perf optimization)
@@ -187,7 +187,7 @@ func (ag *nonReachabilityGrouper) vertexNext() (v1, v2 *Vertex, err error) {
 
 		// if we got here, it means we're skipping over this candidate!
 		if ok, err := ag.baseGrouper.vertexTest(false); err != nil {
-			log.Fatalf("Error running autoGroup(vertexTest): %v", err)
+			log.Fatalf("error running autoGroup(vertexTest): %v", err)
 		} else if !ok {
 			return nil, nil, nil // done!
 		}
@@ -284,7 +284,7 @@ func (g *Graph) VertexMerge(v1, v2 *Vertex, vertexMergeFn func(*Vertex, *Vertex)
 
 	// 5) creation of a cyclic graph should throw an error
 	if _, err := g.TopologicalSort(); err != nil { // am i a dag or not?
-		return errwrap.Wrapf(err, "TopologicalSort failed") // not a dag
+		return errwrap.Wrapf(err, "the TopologicalSort failed") // not a dag
 	}
 	return nil // success
 }
@@ -295,14 +295,14 @@ func (g *Graph) autoGroup(ag AutoGrouper) chan string {
 	go func(strch chan string) {
 		strch <- fmt.Sprintf("Compile: Grouping: Algorithm: %v...", ag.name())
 		if err := ag.init(g); err != nil {
-			log.Fatalf("Error running autoGroup(init): %v", err)
+			log.Fatalf("error running autoGroup(init): %v", err)
 		}
 
 		for {
 			var v, w *Vertex
 			v, w, err := ag.vertexNext() // get pair to compare
 			if err != nil {
-				log.Fatalf("Error running autoGroup(vertexNext): %v", err)
+				log.Fatalf("error running autoGroup(vertexNext): %v", err)
 			}
 			merged := false
 			// save names since they change during the runs
@@ -325,7 +325,7 @@ func (g *Graph) autoGroup(ag AutoGrouper) chan string {
 
 			// did these get used?
 			if ok, err := ag.vertexTest(merged); err != nil {
-				log.Fatalf("Error running autoGroup(vertexTest): %v", err)
+				log.Fatalf("error running autoGroup(vertexTest): %v", err)
 			} else if !ok {
 				break // done!
 			}

pgraph/autogroup_test.go

@@ -0,0 +1,486 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package pgraph
+
+import (
+	"testing"
+)
+
+// all of the following test cases are laid out with the following semantics:
+// * vertices which start with the same single letter are considered "like"
+// * "like" elements should be merged
+// * vertices can have any integer after their single letter "family" type
+// * grouped vertices should have a name with a comma separated list of names
+// * edges follow the same conventions about grouping
+
+// empty graph
+func TestPgraphGrouping1(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	g2 := NewGraph("g2") // expected result
+	runGraphCmp(t, g1, g2)
+}
+
+// single vertex
+func TestPgraphGrouping2(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{                    // grouping to limit variable scope
+		a1 := NewVertex(NewNoopResTest("a1"))
+		g1.AddVertex(a1)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		g2.AddVertex(a1)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// two vertices
+func TestPgraphGrouping3(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		g1.AddVertex(a1, b1)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		g2.AddVertex(a1, b1)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// two vertices merge
+func TestPgraphGrouping4(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		g1.AddVertex(a1, a2)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2"))
+		g2.AddVertex(a)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// three vertices merge
+func TestPgraphGrouping5(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		a3 := NewVertex(NewNoopResTest("a3"))
+		g1.AddVertex(a1, a2, a3)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2,a3"))
+		g2.AddVertex(a)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// three vertices, two merge
+func TestPgraphGrouping6(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		g1.AddVertex(a1, a2, b1)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		g2.AddVertex(a, b1)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// four vertices, three merge
+func TestPgraphGrouping7(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		a3 := NewVertex(NewNoopResTest("a3"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		g1.AddVertex(a1, a2, a3, b1)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2,a3"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		g2.AddVertex(a, b1)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// four vertices, two&two merge
+func TestPgraphGrouping8(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		b2 := NewVertex(NewNoopResTest("b2"))
+		g1.AddVertex(a1, a2, b1, b2)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2"))
+		b := NewVertex(NewNoopResTest("b1,b2"))
+		g2.AddVertex(a, b)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// five vertices, two&three merge
+func TestPgraphGrouping9(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		b2 := NewVertex(NewNoopResTest("b2"))
+		b3 := NewVertex(NewNoopResTest("b3"))
+		g1.AddVertex(a1, a2, b1, b2, b3)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2"))
+		b := NewVertex(NewNoopResTest("b1,b2,b3"))
+		g2.AddVertex(a, b)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// three unique vertices
+func TestPgraphGrouping10(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		g1.AddVertex(a1, b1, c1)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		g2.AddVertex(a1, b1, c1)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// three unique vertices, two merge
+func TestPgraphGrouping11(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		b2 := NewVertex(NewNoopResTest("b2"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		g1.AddVertex(a1, b1, b2, c1)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b := NewVertex(NewNoopResTest("b1,b2"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		g2.AddVertex(a1, b, c1)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// simple merge 1
+// a1   a2         a1,a2
+//   \ /     >>>     |     (arrows point downwards)
+//    b              b
+func TestPgraphGrouping12(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		g1.AddEdge(a1, b1, e1)
+		g1.AddEdge(a2, b1, e2)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		e := NewEdge("e1,e2")
+		g2.AddEdge(a, b1, e)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// simple merge 2
+//    b              b
+//   / \     >>>     |     (arrows point downwards)
+// a1   a2         a1,a2
+func TestPgraphGrouping13(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		g1.AddEdge(b1, a1, e1)
+		g1.AddEdge(b1, a2, e2)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		e := NewEdge("e1,e2")
+		g2.AddEdge(b1, a, e)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// triple merge
+// a1 a2  a3         a1,a2,a3
+//   \ | /     >>>       |      (arrows point downwards)
+//     b                 b
+func TestPgraphGrouping14(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		a3 := NewVertex(NewNoopResTest("a3"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		e3 := NewEdge("e3")
+		g1.AddEdge(a1, b1, e1)
+		g1.AddEdge(a2, b1, e2)
+		g1.AddEdge(a3, b1, e3)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2,a3"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		e := NewEdge("e1,e2,e3")
+		g2.AddEdge(a, b1, e)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// chain merge
+//    a1             a1
+//   /  \             |
+// b1    b2   >>>   b1,b2   (arrows point downwards)
+//   \  /             |
+//    c1             c1
+func TestPgraphGrouping15(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		b2 := NewVertex(NewNoopResTest("b2"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		e3 := NewEdge("e3")
+		e4 := NewEdge("e4")
+		g1.AddEdge(a1, b1, e1)
+		g1.AddEdge(a1, b2, e2)
+		g1.AddEdge(b1, c1, e3)
+		g1.AddEdge(b2, c1, e4)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b := NewVertex(NewNoopResTest("b1,b2"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		e1 := NewEdge("e1,e2")
+		e2 := NewEdge("e3,e4")
+		g2.AddEdge(a1, b, e1)
+		g2.AddEdge(b, c1, e2)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// re-attach 1 (outer)
+// technically the second possibility is valid too, depending on which order we
+// merge edges in, and if we don't filter out any unnecessary edges afterwards!
+// a1    a2         a1,a2        a1,a2
+//  |   /             |            |  \
+// b1  /      >>>    b1     OR    b1  /   (arrows point downwards)
+//  | /               |            | /
+// c1                c1           c1
+func TestPgraphGrouping16(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		e3 := NewEdge("e3")
+		g1.AddEdge(a1, b1, e1)
+		g1.AddEdge(b1, c1, e2)
+		g1.AddEdge(a2, c1, e3)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		e1 := NewEdge("e1,e3")
+		e2 := NewEdge("e2,e3") // e3 gets "merged through" to BOTH edges!
+		g2.AddEdge(a, b1, e1)
+		g2.AddEdge(b1, c1, e2)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// re-attach 2 (inner)
+// a1    b2          a1
+//  |   /             |
+// b1  /      >>>   b1,b2   (arrows point downwards)
+//  | /               |
+// c1                c1
+func TestPgraphGrouping17(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		b2 := NewVertex(NewNoopResTest("b2"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		e3 := NewEdge("e3")
+		g1.AddEdge(a1, b1, e1)
+		g1.AddEdge(b1, c1, e2)
+		g1.AddEdge(b2, c1, e3)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b := NewVertex(NewNoopResTest("b1,b2"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2,e3")
+		g2.AddEdge(a1, b, e1)
+		g2.AddEdge(b, c1, e2)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// re-attach 3 (double)
+// similar to "re-attach 1", technically there is a second possibility for this
+// a2   a1    b2         a1,a2
+//   \   |   /             |
+//    \ b1  /      >>>   b1,b2   (arrows point downwards)
+//     \ | /               |
+//      c1                c1
+func TestPgraphGrouping18(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		b1 := NewVertex(NewNoopResTest("b1"))
+		b2 := NewVertex(NewNoopResTest("b2"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		e3 := NewEdge("e3")
+		e4 := NewEdge("e4")
+		g1.AddEdge(a1, b1, e1)
+		g1.AddEdge(b1, c1, e2)
+		g1.AddEdge(a2, c1, e3)
+		g1.AddEdge(b2, c1, e4)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a := NewVertex(NewNoopResTest("a1,a2"))
+		b := NewVertex(NewNoopResTest("b1,b2"))
+		c1 := NewVertex(NewNoopResTest("c1"))
+		e1 := NewEdge("e1,e3")
+		e2 := NewEdge("e2,e3,e4") // e3 gets "merged through" to BOTH edges!
+		g2.AddEdge(a, b, e1)
+		g2.AddEdge(b, c1, e2)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// connected merge 0, (no change!)
+// a1            a1
+//   \     >>>     \     (arrows point downwards)
+//    a2            a2
+func TestPgraphGroupingConnected0(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		e1 := NewEdge("e1")
+		g1.AddEdge(a1, a2, e1)
+	}
+	g2 := NewGraph("g2") // expected result ?
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		e1 := NewEdge("e1")
+		g2.AddEdge(a1, a2, e1)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+// connected merge 1, (no change!)
+// a1              a1
+//   \               \
+//    b      >>>      b      (arrows point downwards)
+//     \               \
+//      a2              a2
+func TestPgraphGroupingConnected1(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b := NewVertex(NewNoopResTest("b"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		g1.AddEdge(a1, b, e1)
+		g1.AddEdge(b, a2, e2)
+	}
+	g2 := NewGraph("g2") // expected result ?
+	{
+		a1 := NewVertex(NewNoopResTest("a1"))
+		b := NewVertex(NewNoopResTest("b"))
+		a2 := NewVertex(NewNoopResTest("a2"))
+		e1 := NewEdge("e1")
+		e2 := NewEdge("e2")
+		g2.AddEdge(a1, b, e1)
+		g2.AddEdge(b, a2, e2)
+	}
+	runGraphCmp(t, g1, g2)
+}

pgraph/graphviz.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -15,7 +15,7 @@
 // You should have received a copy of the GNU Affero General Public License
 // along with this program.  If not, see <http://www.gnu.org/licenses/>.
 
-package pgraph
+package pgraph // TODO: this should be a subpackage
 
 import (
 	"fmt"
@@ -46,11 +46,15 @@ func (g *Graph) Graphviz() (out string) {
 	//out += "\tnode [shape=box];\n"
 	str := ""
 	for i := range g.Adjacency { // reverse paths
-		out += fmt.Sprintf("\t%s [label=\"%s[%s]\"];\n", i.GetName(), i.Kind(), i.GetName())
+		out += fmt.Sprintf("\t\"%s\" [label=\"%s[%s]\"];\n", i.GetName(), i.Kind(), i.GetName())
 		for j := range g.Adjacency[i] {
 			k := g.Adjacency[i][j]
 			// use str for clearer output ordering
-			str += fmt.Sprintf("\t%s -> %s [label=%s];\n", i.GetName(), j.GetName(), k.Name)
+			if k.Notify {
+				str += fmt.Sprintf("\t\"%s\" -> \"%s\" [label=\"%s\",style=bold];\n", i.GetName(), j.GetName(), k.Name)
+			} else {
+				str += fmt.Sprintf("\t\"%s\" -> \"%s\" [label=\"%s\"];\n", i.GetName(), j.GetName(), k.Name)
+			}
 		}
 	}
 	out += str
@@ -60,16 +64,20 @@ func (g *Graph) Graphviz() (out string) {
 
 // ExecGraphviz writes out the graphviz data and runs the correct graphviz
 // filter command.
-func (g *Graph) ExecGraphviz(program, filename string) error {
+func (g *Graph) ExecGraphviz(program, filename, hostname string) error {
 
 	switch program {
 	case "dot", "neato", "twopi", "circo", "fdp":
 	default:
-		return fmt.Errorf("Invalid graphviz program selected!")
+		return fmt.Errorf("invalid graphviz program selected")
 	}
 
 	if filename == "" {
-		return fmt.Errorf("No filename given!")
+		return fmt.Errorf("no filename given")
+	}
+
+	if hostname != "" {
+		filename = fmt.Sprintf("%s@%s", filename, hostname)
 	}
 
 	// run as a normal user if possible when run with sudo
@@ -78,18 +86,18 @@ func (g *Graph) ExecGraphviz(program, filename string) error {
 
 	err := ioutil.WriteFile(filename, []byte(g.Graphviz()), 0644)
 	if err != nil {
-		return fmt.Errorf("Error writing to filename!")
+		return fmt.Errorf("error writing to filename")
 	}
 
 	if err1 == nil && err2 == nil {
 		if err := os.Chown(filename, uid, gid); err != nil {
-			return fmt.Errorf("Error changing file owner!")
+			return fmt.Errorf("error changing file owner")
 		}
 	}
 
 	path, err := exec.LookPath(program)
 	if err != nil {
-		return fmt.Errorf("Graphviz is missing!")
+		return fmt.Errorf("the Graphviz program is missing")
 	}
 
 	out := fmt.Sprintf("%s.png", filename)
@@ -104,7 +112,7 @@ func (g *Graph) ExecGraphviz(program, filename string) error {
 	}
 	_, err = cmd.Output()
 	if err != nil {
-		return fmt.Errorf("Error writing to image!")
+		return fmt.Errorf("error writing to image")
 	}
 	return nil
 }

pgraph/pgraph.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -24,7 +24,9 @@ import (
 	"sync"
 
 	"github.com/purpleidea/mgmt/event"
+	"github.com/purpleidea/mgmt/prometheus"
 	"github.com/purpleidea/mgmt/resources"
+	"github.com/purpleidea/mgmt/util/semaphore"
 
 	errwrap "github.com/pkg/errors"
 )
@@ -40,6 +42,7 @@ const (
 	graphStatePaused
 )
 
+// Flags contains specific constants used by the graph.
 type Flags struct {
 	Debug bool
 }
@@ -55,7 +58,13 @@ type Graph struct {
 	Adjacency map[*Vertex]map[*Vertex]*Edge // *Vertex -> *Vertex (edge)
 	Flags     Flags
 	state     graphState
-	mutex     sync.Mutex // used when modifying graph State variable
+	fastPause bool        // used to disable pokes for a fast pause
+	mutex     *sync.Mutex // used when modifying graph State variable
+	wg        *sync.WaitGroup
+	semas     map[string]*semaphore.Semaphore
+	slock     *sync.Mutex // semaphore mutex
+
+	prometheus *prometheus.Prometheus // the prometheus instance
 }
 
 // Vertex is the primary vertex struct in this library.
@@ -78,6 +87,11 @@ func NewGraph(name string) *Graph {
 		Name:      name,
 		Adjacency: make(map[*Vertex]map[*Vertex]*Edge),
 		state:     graphStateNil,
+		// ptr b/c: Mutex/WaitGroup must not be copied after first use
+		mutex: &sync.Mutex{},
+		wg:    &sync.WaitGroup{},
+		semas: make(map[string]*semaphore.Semaphore),
+		slock: &sync.Mutex{},
 	}
 }
 
@@ -112,6 +126,13 @@ func (g *Graph) Copy() *Graph {
 		Adjacency: make(map[*Vertex]map[*Vertex]*Edge, len(g.Adjacency)),
 		Flags:     g.Flags,
 		state:     g.state,
+		mutex:     g.mutex,
+		wg:        g.wg,
+		semas:     g.semas,
+		slock:     g.slock,
+		fastPause: g.fastPause,
+
+		prometheus: g.prometheus,
 	}
 	for k, v := range g.Adjacency {
 		newGraph.Adjacency[k] = v // copy
@@ -185,17 +206,28 @@ func (g *Graph) DeleteEdge(e *Edge) {
 	}
 }
 
-// GetVertexMatch searches for an equivalent resource in the graph and returns
-// the vertex it is found in, or nil if not found.
-func (g *Graph) GetVertexMatch(obj resources.Res) *Vertex {
-	for k := range g.Adjacency {
-		if k.Res.Compare(obj) {
-			return k
+// CompareMatch searches for an equivalent resource in the graph and returns the
+// vertex it is found in, or nil if not found.
+func (g *Graph) CompareMatch(obj resources.Res) *Vertex {
+	for v := range g.Adjacency {
+		if v.Res.Compare(obj) {
+			return v
 		}
 	}
 	return nil
 }
 
+// TODO: consider adding a mutate API.
+//func (g *Graph) MutateMatch(obj resources.Res) *Vertex {
+//	for v := range g.Adjacency {
+//		if err := v.Res.Mutate(obj); err == nil {
+//			// transmogrified!
+//			return v
+//		}
+//	}
+//	return nil
+//}
+
 // HasVertex returns if the input vertex exists in the graph.
 func (g *Graph) HasVertex(v *Vertex) bool {
 	if _, exists := g.Adjacency[v]; exists {
@@ -475,7 +507,7 @@ func (g *Graph) TopologicalSort() ([]*Vertex, error) { // kahn's algorithm
 		if in > 0 {
 			for n := range g.Adjacency[c] {
 				if remaining[n] > 0 {
-					return nil, fmt.Errorf("Not a dag!")
+					return nil, fmt.Errorf("not a dag")
 				}
 			}
 		}
@@ -525,7 +557,8 @@ func (g *Graph) Reachability(a, b *Vertex) []*Vertex {
 }
 
 // GraphSync updates the oldGraph so that it matches the newGraph receiver. It
-// leaves identical elements alone so that they don't need to be refreshed.
+// leaves identical elements alone so that they don't need to be refreshed. It
+// tries to mutate existing elements into new ones, if they support this.
 // FIXME: add test cases
 func (g *Graph) GraphSync(oldGraph *Graph) (*Graph, error) {
 
@@ -540,13 +573,24 @@ func (g *Graph) GraphSync(oldGraph *Graph) (*Graph, error) {
 
 	for v := range g.Adjacency { // loop through the vertices (resources)
 		res := v.Res // resource
+		var vertex *Vertex
 
-		vertex := oldGraph.GetVertexMatch(res)
-		if vertex == nil { // no match found
-			if err := res.Init(); err != nil {
-				return nil, errwrap.Wrapf(err, "could not Init() resource")
+		// step one, direct compare with res.Compare
+		if vertex == nil { // redundant guard for consistency
+			vertex = oldGraph.CompareMatch(res)
+		}
+
+		// TODO: consider adding a mutate API.
+		// step two, try and mutate with res.Mutate
+		//if vertex == nil { // not found yet...
+		//	vertex = oldGraph.MutateMatch(res)
+		//}
+
+		if vertex == nil { // no match found yet
+			if err := res.Validate(); err != nil {
+				return nil, errwrap.Wrapf(err, "could not Validate() resource")
 			}
-			vertex = NewVertex(res)
+			vertex = v
 			oldGraph.AddVertex(vertex) // call standalone in case not part of an edge
 		}
 		lookup[v] = vertex                      // used for constructing edges
@@ -557,7 +601,8 @@ func (g *Graph) GraphSync(oldGraph *Graph) (*Graph, error) {
 	for v := range oldGraph.Adjacency {
 		if !VertexContains(v, vertexKeep) {
 			// wait for exit before starting new graph!
-			v.SendEvent(event.EventExit, true, false)
+			v.SendEvent(event.EventExit, nil) // sync
+			v.Res.WaitGroup().Wait()
 			oldGraph.DeleteVertex(v)
 		}
 	}
@@ -570,13 +615,13 @@ func (g *Graph) GraphSync(oldGraph *Graph) (*Graph, error) {
 			// lookup vertices (these should exist now)
 			//res1 := v1.Res // resource
 			//res2 := v2.Res
-			//vertex1 := oldGraph.GetVertexMatch(res1)
-			//vertex2 := oldGraph.GetVertexMatch(res2)
+			//vertex1 := oldGraph.CompareMatch(res1)
+			//vertex2 := oldGraph.CompareMatch(res2)
 			vertex1, exists1 := lookup[v1]
 			vertex2, exists2 := lookup[v2]
 			if !exists1 || !exists2 { // no match found, bug?
 				//if vertex1 == nil || vertex2 == nil { // no match found
-				return nil, fmt.Errorf("New vertices weren't found!") // programming error
+				return nil, fmt.Errorf("new vertices weren't found") // programming error
 			}
 
 			edge, exists := oldGraph.Adjacency[vertex1][vertex2]
@@ -614,8 +659,11 @@ func (g *Graph) GraphMetas() []*resources.MetaParams {
 
 // AssociateData associates some data with the object in the graph in question.
 func (g *Graph) AssociateData(data *resources.Data) {
+	// prometheus needs to be associated to this graph as well
+	g.prometheus = data.Prometheus
+
 	for k := range g.Adjacency {
-		k.Res.AssociateData(data)
+		*k.Res.Data() = *data
 	}
 }
 

pgraph/pgraph_test.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -24,13 +24,18 @@ import (
 	"strings"
 	"testing"
 	"time"
+
+	"github.com/purpleidea/mgmt/resources"
+	"github.com/purpleidea/mgmt/util"
 )
 
 // NV is a helper function to make testing easier. It creates a new noop vertex.
 func NV(s string) *Vertex {
-	obj, err := NewNoopRes(s)
-	if err != nil {
-		panic(err) // unlikely test failure!
+	obj := &resources.NoopRes{
+		BaseRes: resources.BaseRes{
+			Name: s,
+		},
+		Comment: "Testing!",
 	}
 	return NewVertex(obj)
 }
@@ -40,11 +45,11 @@ func TestPgraphT1(t *testing.T) {
 	G := NewGraph("g1")
 
 	if i := G.NumVertices(); i != 0 {
-		t.Errorf("Should have 0 vertices instead of: %d.", i)
+		t.Errorf("should have 0 vertices instead of: %d", i)
 	}
 
 	if i := G.NumEdges(); i != 0 {
-		t.Errorf("Should have 0 edges instead of: %d.", i)
+		t.Errorf("should have 0 edges instead of: %d", i)
 	}
 
 	v1 := NV("v1")
@@ -53,11 +58,11 @@ func TestPgraphT1(t *testing.T) {
 	G.AddEdge(v1, v2, e1)
 
 	if i := G.NumVertices(); i != 2 {
-		t.Errorf("Should have 2 vertices instead of: %d.", i)
+		t.Errorf("should have 2 vertices instead of: %d", i)
 	}
 
 	if i := G.NumEdges(); i != 1 {
-		t.Errorf("Should have 1 edges instead of: %d.", i)
+		t.Errorf("should have 1 edges instead of: %d", i)
 	}
 }
 
@@ -84,7 +89,7 @@ func TestPgraphT2(t *testing.T) {
 	G.AddEdge(v5, v6, e5)
 
 	if i := G.NumVertices(); i != 6 {
-		t.Errorf("Should have 6 vertices instead of: %d.", i)
+		t.Errorf("should have 6 vertices instead of: %d", i)
 	}
 }
 
@@ -112,19 +117,19 @@ func TestPgraphT3(t *testing.T) {
 	//G.AddEdge(v6, v4, e6)
 	out1 := G.DFS(v1)
 	if i := len(out1); i != 3 {
-		t.Errorf("Should have 3 vertices instead of: %d.", i)
-		t.Errorf("Found: %v", out1)
+		t.Errorf("should have 3 vertices instead of: %d", i)
+		t.Errorf("found: %v", out1)
 		for _, v := range out1 {
-			t.Errorf("Value: %v", v.GetName())
+			t.Errorf("value: %v", v.GetName())
 		}
 	}
 
 	out2 := G.DFS(v4)
 	if i := len(out2); i != 3 {
-		t.Errorf("Should have 3 vertices instead of: %d.", i)
-		t.Errorf("Found: %v", out1)
+		t.Errorf("should have 3 vertices instead of: %d", i)
+		t.Errorf("found: %v", out1)
 		for _, v := range out1 {
-			t.Errorf("Value: %v", v.GetName())
+			t.Errorf("value: %v", v.GetName())
 		}
 	}
 }
@@ -144,10 +149,10 @@ func TestPgraphT4(t *testing.T) {
 
 	out := G.DFS(v1)
 	if i := len(out); i != 3 {
-		t.Errorf("Should have 3 vertices instead of: %d.", i)
-		t.Errorf("Found: %v", out)
+		t.Errorf("should have 3 vertices instead of: %d", i)
+		t.Errorf("found: %v", out)
 		for _, v := range out {
-			t.Errorf("Value: %v", v.GetName())
+			t.Errorf("value: %v", v.GetName())
 		}
 	}
 }
@@ -177,7 +182,7 @@ func TestPgraphT5(t *testing.T) {
 	save := []*Vertex{v1, v2, v3}
 	out := G.FilterGraph("new g5", save)
 	if i := out.NumVertices(); i != 3 {
-		t.Errorf("Should have 3 vertices instead of: %d.", i)
+		t.Errorf("should have 3 vertices instead of: %d", i)
 	}
 }
 
@@ -214,7 +219,7 @@ func TestPgraphT6(t *testing.T) {
 	}
 
 	if i := HeisenbergGraphCount(graphs); i != 2 {
-		t.Errorf("Should have 2 graphs instead of: %d.", i)
+		t.Errorf("should have 2 graphs instead of: %d", i)
 	}
 }
 
@@ -232,31 +237,31 @@ func TestPgraphT7(t *testing.T) {
 	G.AddEdge(v3, v1, e3)
 
 	if i := G.NumVertices(); i != 3 {
-		t.Errorf("Should have 3 vertices instead of: %d.", i)
+		t.Errorf("should have 3 vertices instead of: %d", i)
 	}
 
 	G.DeleteVertex(v2)
 
 	if i := G.NumVertices(); i != 2 {
-		t.Errorf("Should have 2 vertices instead of: %d.", i)
+		t.Errorf("should have 2 vertices instead of: %d", i)
 	}
 
 	G.DeleteVertex(v1)
 
 	if i := G.NumVertices(); i != 1 {
-		t.Errorf("Should have 1 vertices instead of: %d.", i)
+		t.Errorf("should have 1 vertices instead of: %d", i)
 	}
 
 	G.DeleteVertex(v3)
 
 	if i := G.NumVertices(); i != 0 {
-		t.Errorf("Should have 0 vertices instead of: %d.", i)
+		t.Errorf("should have 0 vertices instead of: %d", i)
 	}
 
 	G.DeleteVertex(v2) // duplicate deletes don't error...
 
 	if i := G.NumVertices(); i != 0 {
-		t.Errorf("Should have 0 vertices instead of: %d.", i)
+		t.Errorf("should have 0 vertices instead of: %d", i)
 	}
 }
 
@@ -266,26 +271,26 @@ func TestPgraphT8(t *testing.T) {
 	v2 := NV("v2")
 	v3 := NV("v3")
 	if VertexContains(v1, []*Vertex{v1, v2, v3}) != true {
-		t.Errorf("Should be true instead of false.")
+		t.Errorf("should be true instead of false.")
 	}
 
 	v4 := NV("v4")
 	v5 := NV("v5")
 	v6 := NV("v6")
 	if VertexContains(v4, []*Vertex{v5, v6}) != false {
-		t.Errorf("Should be false instead of true.")
+		t.Errorf("should be false instead of true.")
 	}
 
 	v7 := NV("v7")
 	v8 := NV("v8")
 	v9 := NV("v9")
 	if VertexContains(v8, []*Vertex{v7, v8, v9}) != true {
-		t.Errorf("Should be true instead of false.")
+		t.Errorf("should be true instead of false.")
 	}
 
 	v1b := NV("v1") // same value, different objects
 	if VertexContains(v1b, []*Vertex{v1, v2, v3}) != false {
-		t.Errorf("Should be false instead of true.")
+		t.Errorf("should be false instead of true.")
 	}
 }
 
@@ -314,49 +319,49 @@ func TestPgraphT9(t *testing.T) {
 
 	indegree := G.InDegree() // map[*Vertex]int
 	if i := indegree[v1]; i != 0 {
-		t.Errorf("Indegree of v1 should be 0 instead of: %d.", i)
+		t.Errorf("indegree of v1 should be 0 instead of: %d", i)
 	}
 	if i := indegree[v2]; i != 1 {
-		t.Errorf("Indegree of v2 should be 1 instead of: %d.", i)
+		t.Errorf("indegree of v2 should be 1 instead of: %d", i)
 	}
 	if i := indegree[v3]; i != 1 {
-		t.Errorf("Indegree of v3 should be 1 instead of: %d.", i)
+		t.Errorf("indegree of v3 should be 1 instead of: %d", i)
 	}
 	if i := indegree[v4]; i != 2 {
-		t.Errorf("Indegree of v4 should be 2 instead of: %d.", i)
+		t.Errorf("indegree of v4 should be 2 instead of: %d", i)
 	}
 	if i := indegree[v5]; i != 1 {
-		t.Errorf("Indegree of v5 should be 1 instead of: %d.", i)
+		t.Errorf("indegree of v5 should be 1 instead of: %d", i)
 	}
 	if i := indegree[v6]; i != 1 {
-		t.Errorf("Indegree of v6 should be 1 instead of: %d.", i)
+		t.Errorf("indegree of v6 should be 1 instead of: %d", i)
 	}
 
 	outdegree := G.OutDegree() // map[*Vertex]int
 	if i := outdegree[v1]; i != 2 {
-		t.Errorf("Outdegree of v1 should be 2 instead of: %d.", i)
+		t.Errorf("outdegree of v1 should be 2 instead of: %d", i)
 	}
 	if i := outdegree[v2]; i != 1 {
-		t.Errorf("Outdegree of v2 should be 1 instead of: %d.", i)
+		t.Errorf("outdegree of v2 should be 1 instead of: %d", i)
 	}
 	if i := outdegree[v3]; i != 1 {
-		t.Errorf("Outdegree of v3 should be 1 instead of: %d.", i)
+		t.Errorf("outdegree of v3 should be 1 instead of: %d", i)
 	}
 	if i := outdegree[v4]; i != 1 {
-		t.Errorf("Outdegree of v4 should be 1 instead of: %d.", i)
+		t.Errorf("outdegree of v4 should be 1 instead of: %d", i)
 	}
 	if i := outdegree[v5]; i != 1 {
-		t.Errorf("Outdegree of v5 should be 1 instead of: %d.", i)
+		t.Errorf("outdegree of v5 should be 1 instead of: %d", i)
 	}
 	if i := outdegree[v6]; i != 0 {
-		t.Errorf("Outdegree of v6 should be 0 instead of: %d.", i)
+		t.Errorf("outdegree of v6 should be 0 instead of: %d", i)
 	}
 
 	s, err := G.TopologicalSort()
 	// either possibility is a valid toposort
 	match := reflect.DeepEqual(s, []*Vertex{v1, v2, v3, v4, v5, v6}) || reflect.DeepEqual(s, []*Vertex{v1, v3, v2, v4, v5, v6})
 	if err != nil || !match {
-		t.Errorf("Topological sort failed, error: %v.", err)
+		t.Errorf("topological sort failed, error: %v", err)
 		str := "Found:"
 		for _, v := range s {
 			str += " " + v.Res.GetName()
@@ -388,7 +393,7 @@ func TestPgraphT10(t *testing.T) {
 	G.AddEdge(v4, v2, e6) // cycle
 
 	if _, err := G.TopologicalSort(); err == nil {
-		t.Errorf("Topological sort passed, but graph is cyclic!")
+		t.Errorf("topological sort passed, but graph is cyclic")
 	}
 }
 
@@ -398,7 +403,7 @@ func TestPgraphReachability0(t *testing.T) {
 		G := NewGraph("g")
 		result := G.Reachability(nil, nil)
 		if result != nil {
-			t.Logf("Reachability failed!")
+			t.Logf("reachability failed")
 			str := "Got:"
 			for _, v := range result {
 				str += " " + v.Res.GetName()
@@ -415,7 +420,7 @@ func TestPgraphReachability0(t *testing.T) {
 		expected := []*Vertex{}
 
 		if !reflect.DeepEqual(result, expected) {
-			t.Logf("Reachability failed!")
+			t.Logf("reachability failed")
 			str := "Got:"
 			for _, v := range result {
 				str += " " + v.Res.GetName()
@@ -446,7 +451,7 @@ func TestPgraphReachability0(t *testing.T) {
 		expected := []*Vertex{}
 
 		if !reflect.DeepEqual(result, expected) {
-			t.Logf("Reachability failed!")
+			t.Logf("reachability failed")
 			str := "Got:"
 			for _, v := range result {
 				str += " " + v.Res.GetName()
@@ -481,7 +486,7 @@ func TestPgraphReachability1(t *testing.T) {
 	expected := []*Vertex{v1, v2, v3, v4, v5, v6}
 
 	if !reflect.DeepEqual(result, expected) {
-		t.Logf("Reachability failed!")
+		t.Logf("reachability failed")
 		str := "Got:"
 		for _, v := range result {
 			str += " " + v.Res.GetName()
@@ -518,7 +523,7 @@ func TestPgraphReachability2(t *testing.T) {
 
 	// !xor test
 	if reflect.DeepEqual(result, expected1) == reflect.DeepEqual(result, expected2) {
-		t.Logf("Reachability failed!")
+		t.Logf("reachability failed")
 		str := "Got:"
 		for _, v := range result {
 			str += " " + v.Res.GetName()
@@ -553,7 +558,7 @@ func TestPgraphReachability3(t *testing.T) {
 	expected := []*Vertex{v1, v5, v6}
 
 	if !reflect.DeepEqual(result, expected) {
-		t.Logf("Reachability failed!")
+		t.Logf("reachability failed")
 		str := "Got:"
 		for _, v := range result {
 			str += " " + v.Res.GetName()
@@ -588,7 +593,7 @@ func TestPgraphReachability4(t *testing.T) {
 	expected := []*Vertex{v1, v6}
 
 	if !reflect.DeepEqual(result, expected) {
-		t.Logf("Reachability failed!")
+		t.Logf("reachability failed")
 		str := "Got:"
 		for _, v := range result {
 			str += " " + v.Res.GetName()
@@ -606,27 +611,27 @@ func TestPgraphT11(t *testing.T) {
 	v6 := NV("v6")
 
 	if rev := Reverse([]*Vertex{}); !reflect.DeepEqual(rev, []*Vertex{}) {
-		t.Errorf("Reverse of vertex slice failed.")
+		t.Errorf("reverse of vertex slice failed")
 	}
 
 	if rev := Reverse([]*Vertex{v1}); !reflect.DeepEqual(rev, []*Vertex{v1}) {
-		t.Errorf("Reverse of vertex slice failed.")
+		t.Errorf("reverse of vertex slice failed")
 	}
 
 	if rev := Reverse([]*Vertex{v1, v2, v3, v4, v5, v6}); !reflect.DeepEqual(rev, []*Vertex{v6, v5, v4, v3, v2, v1}) {
-		t.Errorf("Reverse of vertex slice failed.")
+		t.Errorf("reverse of vertex slice failed")
 	}
 
 	if rev := Reverse([]*Vertex{v6, v5, v4, v3, v2, v1}); !reflect.DeepEqual(rev, []*Vertex{v1, v2, v3, v4, v5, v6}) {
-		t.Errorf("Reverse of vertex slice failed.")
+		t.Errorf("reverse of vertex slice failed")
 	}
 }
 
 type NoopResTest struct {
-	NoopRes
+	resources.NoopRes
 }
 
-func (obj *NoopResTest) GroupCmp(r Res) bool {
+func (obj *NoopResTest) GroupCmp(r resources.Res) bool {
 	res, ok := r.(*NoopResTest)
 	if !ok {
 		return false
@@ -643,16 +648,15 @@ func (obj *NoopResTest) GroupCmp(r Res) bool {
 
 func NewNoopResTest(name string) *NoopResTest {
 	obj := &NoopResTest{
-		NoopRes: NoopRes{
-			BaseRes: BaseRes{
+		NoopRes: resources.NoopRes{
+			BaseRes: resources.BaseRes{
 				Name: name,
-				MetaParams: MetaParams{
+				MetaParams: resources.MetaParams{
 					AutoGroup: true, // always autogroup
 				},
 			},
 		},
 	}
-	obj.Init() // optional here in this testing scenario (for now)
 	return obj
 }
 
@@ -680,10 +684,10 @@ func ListStrCmp(a, b []string) bool {
 // It also compares if grouped element groups are identical
 func GraphCmp(g1, g2 *Graph) error {
 	if n1, n2 := g1.NumVertices(), g2.NumVertices(); n1 != n2 {
-		return fmt.Errorf("Graph g1 has %d vertices, while g2 has %d.", n1, n2)
+		return fmt.Errorf("graph g1 has %d vertices, while g2 has %d", n1, n2)
 	}
 	if e1, e2 := g1.NumEdges(), g2.NumEdges(); e1 != e2 {
-		return fmt.Errorf("Graph g1 has %d edges, while g2 has %d.", e1, e2)
+		return fmt.Errorf("graph g1 has %d edges, while g2 has %d", e1, e2)
 	}
 
 	var m = make(map[*Vertex]*Vertex) // g1 to g2 vertex correspondence
@@ -695,7 +699,7 @@ Loop:
 		for _, x1 := range v1.GetGroup() {
 			l1 = append(l1, x1.GetName()) // add my contents
 		}
-		l1 = StrRemoveDuplicatesInList(l1) // remove duplicates
+		l1 = util.StrRemoveDuplicatesInList(l1) // remove duplicates
 		sort.Strings(l1)
 
 		// inner loop
@@ -705,7 +709,7 @@ Loop:
 			for _, x2 := range v2.GetGroup() {
 				l2 = append(l2, x2.GetName())
 			}
-			l2 = StrRemoveDuplicatesInList(l2) // remove duplicates
+			l2 = util.StrRemoveDuplicatesInList(l2) // remove duplicates
 			sort.Strings(l2)
 
 			// does l1 match l2 ?
@@ -714,7 +718,7 @@ Loop:
 				continue Loop
 			}
 		}
-		return fmt.Errorf("Graph g1, has no match in g2 for: %v", v1.GetName())
+		return fmt.Errorf("graph g1, has no match in g2 for: %v", v1.GetName())
 	}
 	// vertices (and groups) match :)
 
@@ -723,7 +727,7 @@ Loop:
 		v2 := m[v1] // lookup in map to get correspondance
 		// g1.Adjacency[v1] corresponds to g2.Adjacency[v2]
 		if e1, e2 := len(g1.Adjacency[v1]), len(g2.Adjacency[v2]); e1 != e2 {
-			return fmt.Errorf("Graph g1, vertex(%v) has %d edges, while g2, vertex(%v) has %d.", v1.GetName(), e1, v2.GetName(), e2)
+			return fmt.Errorf("graph g1, vertex(%v) has %d edges, while g2, vertex(%v) has %d", v1.GetName(), e1, v2.GetName(), e2)
 		}
 
 		for vv1, ee1 := range g1.Adjacency[v1] {
@@ -738,24 +742,36 @@ Loop:
 			for _, x1 := range vv1.GetGroup() {
 				l1 = append(l1, x1.GetName()) // add my contents
 			}
-			l1 = StrRemoveDuplicatesInList(l1) // remove duplicates
+			l1 = util.StrRemoveDuplicatesInList(l1) // remove duplicates
 			sort.Strings(l1)
 
 			l2 := strings.Split(vv2.GetName(), ",")
 			for _, x2 := range vv2.GetGroup() {
 				l2 = append(l2, x2.GetName())
 			}
-			l2 = StrRemoveDuplicatesInList(l2) // remove duplicates
+			l2 = util.StrRemoveDuplicatesInList(l2) // remove duplicates
 			sort.Strings(l2)
 
 			// does l1 match l2 ?
 			if !ListStrCmp(l1, l2) { // cmp!
-				return fmt.Errorf("Graph g1 and g2 don't agree on: %v and %v", vv1.GetName(), vv2.GetName())
+				return fmt.Errorf("graph g1 and g2 don't agree on: %v and %v", vv1.GetName(), vv2.GetName())
 			}
 
 			// check: (2) ee1 == ee2
 			if ee1.Name != ee2.Name {
-				return fmt.Errorf("Graph g1 edge(%v) doesn't match g2 edge(%v)", ee1.Name, ee2.Name)
+				return fmt.Errorf("graph g1 edge(%v) doesn't match g2 edge(%v)", ee1.Name, ee2.Name)
+			}
+		}
+	}
+
+	// check meta parameters
+	for v1 := range g1.Adjacency { // for each vertex in g1
+		for v2 := range g2.Adjacency { // does it match in g2 ?
+			s1, s2 := v1.Meta().Sema, v2.Meta().Sema
+			sort.Strings(s1)
+			sort.Strings(s2)
+			if !reflect.DeepEqual(s1, s2) {
+				return fmt.Errorf("vertex %s and vertex %s have different semaphores", v1.GetName(), v2.GetName())
 			}
 		}
 	}
@@ -782,7 +798,7 @@ func (ag *testGrouper) vertexMerge(v1, v2 *Vertex) (v *Vertex, err error) {
 	for _, n := range obj.GetGroup() {
 		names = append(names, n.GetName()) // add my contents
 	}
-	names = StrRemoveDuplicatesInList(names) // remove duplicates
+	names = util.StrRemoveDuplicatesInList(names) // remove duplicates
 	sort.Strings(names)
 	obj.SetName(strings.Join(names, ","))
 	return // success or fail, and no need to merge the actual vertices!
@@ -793,7 +809,7 @@ func (ag *testGrouper) edgeMerge(e1, e2 *Edge) *Edge {
 	n1 := strings.Split(e1.Name, ",") // load
 	n2 := strings.Split(e2.Name, ",") // load
 	names := append(n1, n2...)
-	names = StrRemoveDuplicatesInList(names) // remove duplicates
+	names = util.StrRemoveDuplicatesInList(names) // remove duplicates
 	sort.Strings(names)
 	return NewEdge(strings.Join(names, ","))
 }
@@ -801,7 +817,11 @@ func (ag *testGrouper) edgeMerge(e1, e2 *Edge) *Edge {
 func (g *Graph) fullPrint() (str string) {
 	str += "\n"
 	for v := range g.Adjacency {
-		str += fmt.Sprintf("* v: %v\n", v.GetName())
+		if semas := v.Meta().Sema; len(semas) > 0 {
+			str += fmt.Sprintf("* v: %v; sema: %v\n", v.GetName(), semas)
+		} else {
+			str += fmt.Sprintf("* v: %v\n", v.GetName())
+		}
 		// TODO: add explicit grouping data?
 	}
 	for v1 := range g.Adjacency {
@@ -827,476 +847,12 @@ func runGraphCmp(t *testing.T, g1, g2 *Graph) {
 	}
 }
 
-// all of the following test cases are laid out with the following semantics:
-// * vertices which start with the same single letter are considered "like"
-// * "like" elements should be merged
-// * vertices can have any integer after their single letter "family" type
-// * grouped vertices should have a name with a comma separated list of names
-// * edges follow the same conventions about grouping
-
-// empty graph
-func TestPgraphGrouping1(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	g2 := NewGraph("g2") // expected result
-	runGraphCmp(t, g1, g2)
-}
-
-// single vertex
-func TestPgraphGrouping2(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{                    // grouping to limit variable scope
-		a1 := NewVertex(NewNoopResTest("a1"))
-		g1.AddVertex(a1)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		g2.AddVertex(a1)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// two vertices
-func TestPgraphGrouping3(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		g1.AddVertex(a1, b1)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		g2.AddVertex(a1, b1)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// two vertices merge
-func TestPgraphGrouping4(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		g1.AddVertex(a1, a2)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2"))
-		g2.AddVertex(a)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// three vertices merge
-func TestPgraphGrouping5(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		a3 := NewVertex(NewNoopResTest("a3"))
-		g1.AddVertex(a1, a2, a3)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2,a3"))
-		g2.AddVertex(a)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// three vertices, two merge
-func TestPgraphGrouping6(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		g1.AddVertex(a1, a2, b1)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		g2.AddVertex(a, b1)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// four vertices, three merge
-func TestPgraphGrouping7(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		a3 := NewVertex(NewNoopResTest("a3"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		g1.AddVertex(a1, a2, a3, b1)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2,a3"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		g2.AddVertex(a, b1)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// four vertices, two&two merge
-func TestPgraphGrouping8(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		b2 := NewVertex(NewNoopResTest("b2"))
-		g1.AddVertex(a1, a2, b1, b2)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2"))
-		b := NewVertex(NewNoopResTest("b1,b2"))
-		g2.AddVertex(a, b)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// five vertices, two&three merge
-func TestPgraphGrouping9(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		b2 := NewVertex(NewNoopResTest("b2"))
-		b3 := NewVertex(NewNoopResTest("b3"))
-		g1.AddVertex(a1, a2, b1, b2, b3)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2"))
-		b := NewVertex(NewNoopResTest("b1,b2,b3"))
-		g2.AddVertex(a, b)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// three unique vertices
-func TestPgraphGrouping10(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		g1.AddVertex(a1, b1, c1)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		g2.AddVertex(a1, b1, c1)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// three unique vertices, two merge
-func TestPgraphGrouping11(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		b2 := NewVertex(NewNoopResTest("b2"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		g1.AddVertex(a1, b1, b2, c1)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b := NewVertex(NewNoopResTest("b1,b2"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		g2.AddVertex(a1, b, c1)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// simple merge 1
-// a1   a2         a1,a2
-//   \ /     >>>     |     (arrows point downwards)
-//    b              b
-func TestPgraphGrouping12(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		g1.AddEdge(a1, b1, e1)
-		g1.AddEdge(a2, b1, e2)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		e := NewEdge("e1,e2")
-		g2.AddEdge(a, b1, e)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// simple merge 2
-//    b              b
-//   / \     >>>     |     (arrows point downwards)
-// a1   a2         a1,a2
-func TestPgraphGrouping13(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		g1.AddEdge(b1, a1, e1)
-		g1.AddEdge(b1, a2, e2)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		e := NewEdge("e1,e2")
-		g2.AddEdge(b1, a, e)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// triple merge
-// a1 a2  a3         a1,a2,a3
-//   \ | /     >>>       |      (arrows point downwards)
-//     b                 b
-func TestPgraphGrouping14(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		a3 := NewVertex(NewNoopResTest("a3"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		e3 := NewEdge("e3")
-		g1.AddEdge(a1, b1, e1)
-		g1.AddEdge(a2, b1, e2)
-		g1.AddEdge(a3, b1, e3)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2,a3"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		e := NewEdge("e1,e2,e3")
-		g2.AddEdge(a, b1, e)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// chain merge
-//    a1             a1
-//   /  \             |
-// b1    b2   >>>   b1,b2   (arrows point downwards)
-//   \  /             |
-//    c1             c1
-func TestPgraphGrouping15(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		b2 := NewVertex(NewNoopResTest("b2"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		e3 := NewEdge("e3")
-		e4 := NewEdge("e4")
-		g1.AddEdge(a1, b1, e1)
-		g1.AddEdge(a1, b2, e2)
-		g1.AddEdge(b1, c1, e3)
-		g1.AddEdge(b2, c1, e4)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b := NewVertex(NewNoopResTest("b1,b2"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		e1 := NewEdge("e1,e2")
-		e2 := NewEdge("e3,e4")
-		g2.AddEdge(a1, b, e1)
-		g2.AddEdge(b, c1, e2)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// re-attach 1 (outer)
-// technically the second possibility is valid too, depending on which order we
-// merge edges in, and if we don't filter out any unnecessary edges afterwards!
-// a1    a2         a1,a2        a1,a2
-//  |   /             |            |  \
-// b1  /      >>>    b1     OR    b1  /   (arrows point downwards)
-//  | /               |            | /
-// c1                c1           c1
-func TestPgraphGrouping16(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		e3 := NewEdge("e3")
-		g1.AddEdge(a1, b1, e1)
-		g1.AddEdge(b1, c1, e2)
-		g1.AddEdge(a2, c1, e3)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		e1 := NewEdge("e1,e3")
-		e2 := NewEdge("e2,e3") // e3 gets "merged through" to BOTH edges!
-		g2.AddEdge(a, b1, e1)
-		g2.AddEdge(b1, c1, e2)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// re-attach 2 (inner)
-// a1    b2          a1
-//  |   /             |
-// b1  /      >>>   b1,b2   (arrows point downwards)
-//  | /               |
-// c1                c1
-func TestPgraphGrouping17(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		b2 := NewVertex(NewNoopResTest("b2"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		e3 := NewEdge("e3")
-		g1.AddEdge(a1, b1, e1)
-		g1.AddEdge(b1, c1, e2)
-		g1.AddEdge(b2, c1, e3)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b := NewVertex(NewNoopResTest("b1,b2"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2,e3")
-		g2.AddEdge(a1, b, e1)
-		g2.AddEdge(b, c1, e2)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// re-attach 3 (double)
-// similar to "re-attach 1", technically there is a second possibility for this
-// a2   a1    b2         a1,a2
-//   \   |   /             |
-//    \ b1  /      >>>   b1,b2   (arrows point downwards)
-//     \ | /               |
-//      c1                c1
-func TestPgraphGrouping18(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		b1 := NewVertex(NewNoopResTest("b1"))
-		b2 := NewVertex(NewNoopResTest("b2"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		e3 := NewEdge("e3")
-		e4 := NewEdge("e4")
-		g1.AddEdge(a1, b1, e1)
-		g1.AddEdge(b1, c1, e2)
-		g1.AddEdge(a2, c1, e3)
-		g1.AddEdge(b2, c1, e4)
-	}
-	g2 := NewGraph("g2") // expected result
-	{
-		a := NewVertex(NewNoopResTest("a1,a2"))
-		b := NewVertex(NewNoopResTest("b1,b2"))
-		c1 := NewVertex(NewNoopResTest("c1"))
-		e1 := NewEdge("e1,e3")
-		e2 := NewEdge("e2,e3,e4") // e3 gets "merged through" to BOTH edges!
-		g2.AddEdge(a, b, e1)
-		g2.AddEdge(b, c1, e2)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// connected merge 0, (no change!)
-// a1            a1
-//   \     >>>     \     (arrows point downwards)
-//    a2            a2
-func TestPgraphGroupingConnected0(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		e1 := NewEdge("e1")
-		g1.AddEdge(a1, a2, e1)
-	}
-	g2 := NewGraph("g2") // expected result ?
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		e1 := NewEdge("e1")
-		g2.AddEdge(a1, a2, e1)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
-// connected merge 1, (no change!)
-// a1              a1
-//   \               \
-//    b      >>>      b      (arrows point downwards)
-//     \               \
-//      a2              a2
-func TestPgraphGroupingConnected1(t *testing.T) {
-	g1 := NewGraph("g1") // original graph
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b := NewVertex(NewNoopResTest("b"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		g1.AddEdge(a1, b, e1)
-		g1.AddEdge(b, a2, e2)
-	}
-	g2 := NewGraph("g2") // expected result ?
-	{
-		a1 := NewVertex(NewNoopResTest("a1"))
-		b := NewVertex(NewNoopResTest("b"))
-		a2 := NewVertex(NewNoopResTest("a2"))
-		e1 := NewEdge("e1")
-		e2 := NewEdge("e2")
-		g2.AddEdge(a1, b, e1)
-		g2.AddEdge(b, a2, e2)
-	}
-	runGraphCmp(t, g1, g2)
-}
-
 func TestDurationAssumptions(t *testing.T) {
 	var d time.Duration
 	if (d == 0) != true {
-		t.Errorf("Empty time.Duration is no longer equal to zero!")
+		t.Errorf("empty time.Duration is no longer equal to zero")
 	}
 	if (d > 0) != false {
-		t.Errorf("Empty time.Duration is now greater than zero!")
+		t.Errorf("empty time.Duration is now greater than zero")
 	}
 }

pgraph/semaphore.go

@@ -0,0 +1,85 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package pgraph
+
+import (
+	"fmt"
+	"sort"
+	"strconv"
+	"strings"
+
+	"github.com/purpleidea/mgmt/util/semaphore"
+
+	multierr "github.com/hashicorp/go-multierror"
+)
+
+// SemaSep is the trailing separator to split the semaphore id from the size.
+const SemaSep = ":"
+
+// SemaLock acquires the list of semaphores in the graph.
+func (g *Graph) SemaLock(semas []string) error {
+	var reterr error
+	sort.Strings(semas) // very important to avoid deadlock in the dag!
+	for _, id := range semas {
+		g.slock.Lock()          // semaphore creation lock
+		sema, ok := g.semas[id] // lookup
+		if !ok {
+			size := SemaSize(id) // defaults to 1
+			g.semas[id] = semaphore.NewSemaphore(size)
+			sema = g.semas[id]
+		}
+		g.slock.Unlock()
+
+		if err := sema.P(1); err != nil { // lock!
+			reterr = multierr.Append(reterr, err) // list of errors
+		}
+	}
+	return reterr
+}
+
+// SemaUnlock releases the list of semaphores in the graph.
+func (g *Graph) SemaUnlock(semas []string) error {
+	var reterr error
+	sort.Strings(semas) // unlock in the same order to remove partial locks
+	for _, id := range semas {
+		sema, ok := g.semas[id] // lookup
+		if !ok {
+			// programming error!
+			panic(fmt.Sprintf("graph: sema: %s does not exist", id))
+		}
+
+		if err := sema.V(1); err != nil { // unlock!
+			reterr = multierr.Append(reterr, err) // list of errors
+		}
+	}
+	return reterr
+}
+
+// SemaSize returns the size integer associated with the semaphore id. It
+// defaults to 1 if not found.
+func SemaSize(id string) int {
+	size := 1 // default semaphore size
+	// valid id's include "some_id", "hello:42" and ":13"
+	if index := strings.LastIndex(id, SemaSep); index > -1 && (len(id)-index+len(SemaSep)) >= 1 {
+		// NOTE: we only allow size > 0 here!
+		if i, err := strconv.Atoi(id[index+len(SemaSep):]); err == nil && i > 0 {
+			size = i
+		}
+	}
+	return size
+}

pgraph/semaphore_test.go

@@ -0,0 +1,106 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package pgraph
+
+import (
+	"testing"
+
+	"github.com/purpleidea/mgmt/resources"
+)
+
+func TestSemaSize(t *testing.T) {
+	pairs := map[string]int{
+		"id:42":   42,
+		":13":     13,
+		"some_id": 1,
+	}
+	for id, size := range pairs {
+		if i := SemaSize(id); i != size {
+			t.Errorf("sema id `%s`, expected: `%d`, got: `%d`", id, size, i)
+		}
+	}
+}
+
+func NewNoopResTestSema(name string, semas []string) *NoopResTest {
+	obj := &NoopResTest{
+		NoopRes: resources.NoopRes{
+			BaseRes: resources.BaseRes{
+				Name: name,
+				MetaParams: resources.MetaParams{
+					AutoGroup: true, // always autogroup
+					Sema:      semas,
+				},
+			},
+		},
+	}
+	return obj
+}
+
+func TestPgraphSemaphoreGrouping1(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTestSema("a1", []string{"s:1"}))
+		a2 := NewVertex(NewNoopResTestSema("a2", []string{"s:2"}))
+		a3 := NewVertex(NewNoopResTestSema("a3", []string{"s:3"}))
+		g1.AddVertex(a1)
+		g1.AddVertex(a2)
+		g1.AddVertex(a3)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a123 := NewVertex(NewNoopResTestSema("a1,a2,a3", []string{"s:1", "s:2", "s:3"}))
+		g2.AddVertex(a123)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+func TestPgraphSemaphoreGrouping2(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTestSema("a1", []string{"s:10", "s:11"}))
+		a2 := NewVertex(NewNoopResTestSema("a2", []string{"s:2"}))
+		a3 := NewVertex(NewNoopResTestSema("a3", []string{"s:3"}))
+		g1.AddVertex(a1)
+		g1.AddVertex(a2)
+		g1.AddVertex(a3)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a123 := NewVertex(NewNoopResTestSema("a1,a2,a3", []string{"s:10", "s:11", "s:2", "s:3"}))
+		g2.AddVertex(a123)
+	}
+	runGraphCmp(t, g1, g2)
+}
+
+func TestPgraphSemaphoreGrouping3(t *testing.T) {
+	g1 := NewGraph("g1") // original graph
+	{
+		a1 := NewVertex(NewNoopResTestSema("a1", []string{"s:1", "s:2"}))
+		a2 := NewVertex(NewNoopResTestSema("a2", []string{"s:2"}))
+		a3 := NewVertex(NewNoopResTestSema("a3", []string{"s:3"}))
+		g1.AddVertex(a1)
+		g1.AddVertex(a2)
+		g1.AddVertex(a3)
+	}
+	g2 := NewGraph("g2") // expected result
+	{
+		a123 := NewVertex(NewNoopResTestSema("a1,a2,a3", []string{"s:1", "s:2", "s:3"}))
+		g2.AddVertex(a123)
+	}
+	runGraphCmp(t, g1, g2)
+}

prometheus/prometheus.go

@@ -0,0 +1,263 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+// Package prometheus provides functions that are useful to control and manage
+// the build-in prometheus instance.
+package prometheus
+
+import (
+	"errors"
+	"net/http"
+	"strconv"
+	"sync"
+
+	"github.com/prometheus/client_golang/prometheus"
+	"github.com/prometheus/client_golang/prometheus/promhttp"
+
+	errwrap "github.com/pkg/errors"
+)
+
+// DefaultPrometheusListen is registered in
+// https://github.com/prometheus/prometheus/wiki/Default-port-allocations
+const DefaultPrometheusListen = "127.0.0.1:9233"
+
+// ResState represents the status of a resource.
+type ResState int
+
+const (
+	// ResStateOK represents a working resource
+	ResStateOK ResState = iota
+	// ResStateSoftFail represents a resource in soft fail (will be retried)
+	ResStateSoftFail
+	// ResStateHardFail represents a resource in hard fail (will NOT be retried)
+	ResStateHardFail
+)
+
+// Prometheus is the struct that contains information about the
+// prometheus instance. Run Init() on it.
+type Prometheus struct {
+	Listen string // the listen specification for the net/http server
+
+	checkApplyTotal        *prometheus.CounterVec // total of CheckApplies that have been triggered
+	pgraphStartTimeSeconds prometheus.Gauge       // process start time in seconds since unix epoch
+	managedResources       *prometheus.GaugeVec   // Resources we manage now
+	failedResourcesTotal   *prometheus.CounterVec // Total of failures since mgmt has started
+	failedResources        *prometheus.GaugeVec   // Number of current resources
+
+	resourcesState map[string]resStateWithKind // Maps the resources with their current kind/state
+	mutex          *sync.Mutex                 // Mutex used to update resourcesState
+}
+
+// resStateWithKind is used to count the failures by kind
+type resStateWithKind struct {
+	state ResState
+	kind  string
+}
+
+// Init some parameters - currently the Listen address.
+func (obj *Prometheus) Init() error {
+	if len(obj.Listen) == 0 {
+		obj.Listen = DefaultPrometheusListen
+	}
+
+	obj.mutex = &sync.Mutex{}
+	obj.resourcesState = make(map[string]resStateWithKind)
+
+	obj.checkApplyTotal = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Name: "mgmt_checkapply_total",
+			Help: "Number of CheckApply that have run.",
+		},
+		// Labels for this metric.
+		// kind: resource type: Svc, File, ...
+		// apply: if the CheckApply happened in "apply" mode
+		// eventful: did the CheckApply generate an event
+		// errorful: did the CheckApply generate an error
+		[]string{"kind", "apply", "eventful", "errorful"},
+	)
+	prometheus.MustRegister(obj.checkApplyTotal)
+
+	obj.pgraphStartTimeSeconds = prometheus.NewGauge(
+		prometheus.GaugeOpts{
+			Name: "mgmt_graph_start_time_seconds",
+			Help: "Start time of the current graph since unix epoch in seconds.",
+		},
+	)
+	prometheus.MustRegister(obj.pgraphStartTimeSeconds)
+
+	obj.managedResources = prometheus.NewGaugeVec(
+		prometheus.GaugeOpts{
+			Name: "mgmt_resources",
+			Help: "Number of managed resources.",
+		},
+		// kind: resource type: Svc, File, ...
+		[]string{"kind"},
+	)
+	prometheus.MustRegister(obj.managedResources)
+
+	obj.failedResourcesTotal = prometheus.NewCounterVec(
+		prometheus.CounterOpts{
+			Name: "mgmt_failures_total",
+			Help: "Total of failed resources.",
+		},
+		// kind: resource type: Svc, File, ...
+		// failure: soft or hard
+		[]string{"kind", "failure"},
+	)
+	prometheus.MustRegister(obj.failedResourcesTotal)
+
+	obj.failedResources = prometheus.NewGaugeVec(
+		prometheus.GaugeOpts{
+			Name: "mgmt_failures",
+			Help: "Number of failing resources.",
+		},
+		// kind: resource type: Svc, File, ...
+		// failure: soft or hard
+		[]string{"kind", "failure"},
+	)
+	prometheus.MustRegister(obj.failedResources)
+
+	return nil
+}
+
+// Start runs a http server in a go routine, that responds to /metrics
+// as prometheus would expect.
+func (obj *Prometheus) Start() error {
+	http.Handle("/metrics", promhttp.Handler())
+	go http.ListenAndServe(obj.Listen, nil)
+	return nil
+}
+
+// Stop the http server.
+func (obj *Prometheus) Stop() error {
+	// TODO: There is no way in go < 1.8 to stop a http server.
+	// https://stackoverflow.com/questions/39320025/go-how-to-stop-http-listenandserve/41433555#41433555
+	return nil
+}
+
+// UpdateCheckApplyTotal refreshes the failing gauge by parsing the internal
+// state map.
+func (obj *Prometheus) UpdateCheckApplyTotal(kind string, apply, eventful, errorful bool) error {
+	if obj == nil {
+		return nil // happens when mgmt is launched without --prometheus
+	}
+	labels := prometheus.Labels{"kind": kind, "apply": strconv.FormatBool(apply), "eventful": strconv.FormatBool(eventful), "errorful": strconv.FormatBool(errorful)}
+	metric := obj.checkApplyTotal.With(labels)
+	metric.Inc()
+	return nil
+}
+
+// UpdatePgraphStartTime updates the mgmt_graph_start_time_seconds metric
+// to the current timestamp.
+func (obj *Prometheus) UpdatePgraphStartTime() error {
+	if obj == nil {
+		return nil // happens when mgmt is launched without --prometheus
+	}
+	obj.pgraphStartTimeSeconds.SetToCurrentTime()
+	return nil
+}
+
+// AddManagedResource increments the Managed Resource counter and updates the resource status.
+func (obj *Prometheus) AddManagedResource(resUUID string, rtype string) error {
+	if obj == nil {
+		return nil // happens when mgmt is launched without --prometheus
+	}
+	obj.managedResources.With(prometheus.Labels{"kind": rtype}).Inc()
+	if err := obj.UpdateState(resUUID, rtype, ResStateOK); err != nil {
+		return errwrap.Wrapf(err, "can't update the resource status in the map")
+	}
+	return nil
+}
+
+// RemoveManagedResource decrements the Managed Resource counter and updates the resource status.
+func (obj *Prometheus) RemoveManagedResource(resUUID string, rtype string) error {
+	if obj == nil {
+		return nil // happens when mgmt is launched without --prometheus
+	}
+	obj.managedResources.With(prometheus.Labels{"kind": rtype}).Dec()
+	if err := obj.deleteState(resUUID); err != nil {
+		return errwrap.Wrapf(err, "can't remove the resource status from the map")
+	}
+	return nil
+}
+
+// deleteState removes the resources for the state map and re-populates the failing gauge.
+func (obj *Prometheus) deleteState(resUUID string) error {
+	if obj == nil {
+		return nil // happens when mgmt is launched without --prometheus
+	}
+	obj.mutex.Lock()
+	delete(obj.resourcesState, resUUID)
+	obj.mutex.Unlock()
+	if err := obj.updateFailingGauge(); err != nil {
+		return errwrap.Wrapf(err, "can't update the failing gauge")
+	}
+	return nil
+}
+
+// UpdateState updates the state of the resources in our internal state map
+// then triggers a refresh of the failing gauge.
+func (obj *Prometheus) UpdateState(resUUID string, rtype string, newState ResState) error {
+	defer obj.updateFailingGauge()
+	if obj == nil {
+		return nil // happens when mgmt is launched without --prometheus
+	}
+	obj.mutex.Lock()
+	obj.resourcesState[resUUID] = resStateWithKind{state: newState, kind: rtype}
+	obj.mutex.Unlock()
+	if newState != ResStateOK {
+		var strState string
+		if newState == ResStateSoftFail {
+			strState = "soft"
+		} else if newState == ResStateHardFail {
+			strState = "hard"
+		} else {
+			return errors.New("state should be soft or hard failure")
+		}
+		obj.failedResourcesTotal.With(prometheus.Labels{"kind": rtype, "failure": strState}).Inc()
+	}
+	return nil
+}
+
+// updateFailingGauge refreshes the failing gauge by parsking the internal
+// state map.
+func (obj *Prometheus) updateFailingGauge() error {
+	if obj == nil {
+		return nil // happens when mgmt is launched without --prometheus
+	}
+	var softFails, hardFails map[string]float64
+	softFails = make(map[string]float64)
+	hardFails = make(map[string]float64)
+	for _, v := range obj.resourcesState {
+		if v.state == ResStateSoftFail {
+			softFails[v.kind]++
+		} else if v.state == ResStateHardFail {
+			hardFails[v.kind]++
+		}
+	}
+	// TODO: we might want to Zero the metrics we are not using
+	// because in prometheus design the metrics keep living for some time
+	// even after they are removed.
+	obj.failedResources.Reset()
+	for k, v := range softFails {
+		obj.failedResources.With(prometheus.Labels{"kind": k, "failure": "soft"}).Set(v)
+	}
+	for k, v := range hardFails {
+		obj.failedResources.With(prometheus.Labels{"kind": k, "failure": "hard"}).Set(v)
+	}
+	return nil
+}

puppet/gapi.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -50,10 +50,10 @@ func NewGAPI(data gapi.Data, puppetParam *string, puppetConf string) (*GAPI, err
 // Init initializes the puppet GAPI struct.
 func (obj *GAPI) Init(data gapi.Data) error {
 	if obj.initialized {
-		return fmt.Errorf("Already initialized!")
+		return fmt.Errorf("already initialized")
 	}
 	if obj.PuppetParam == nil {
-		return fmt.Errorf("The PuppetParam param must be specified!")
+		return fmt.Errorf("the PuppetParam param must be specified")
 	}
 	obj.data = data // store for later
 	obj.closeChan = make(chan struct{})
@@ -64,23 +64,23 @@ func (obj *GAPI) Init(data gapi.Data) error {
 // Graph returns a current Graph.
 func (obj *GAPI) Graph() (*pgraph.Graph, error) {
 	if !obj.initialized {
-		return nil, fmt.Errorf("Puppet: GAPI is not initialized!")
+		return nil, fmt.Errorf("the puppet GAPI is not initialized")
 	}
 	config := ParseConfigFromPuppet(*obj.PuppetParam, obj.PuppetConf)
 	if config == nil {
-		return nil, fmt.Errorf("Puppet: ParseConfigFromPuppet returned nil!")
+		return nil, fmt.Errorf("function ParseConfigFromPuppet returned nil")
 	}
 	g, err := config.NewGraphFromConfig(obj.data.Hostname, obj.data.World, obj.data.Noop)
 	return g, err
 }
 
-// SwitchStream returns nil errors every time there could be a new graph.
-func (obj *GAPI) SwitchStream() chan error {
+// Next returns nil errors every time there could be a new graph.
+func (obj *GAPI) Next() chan error {
 	if obj.data.NoWatch {
 		return nil
 	}
 	puppetChan := func() <-chan time.Time { // helper function
-		return time.Tick(time.Duration(PuppetInterval(obj.PuppetConf)) * time.Second)
+		return time.Tick(time.Duration(RefreshInterval(obj.PuppetConf)) * time.Second)
 	}
 	ch := make(chan error)
 	obj.wg.Add(1)
@@ -88,7 +88,7 @@ func (obj *GAPI) SwitchStream() chan error {
 		defer obj.wg.Done()
 		defer close(ch) // this will run before the obj.wg.Done()
 		if !obj.initialized {
-			ch <- fmt.Errorf("Puppet: GAPI is not initialized!")
+			ch <- fmt.Errorf("the puppet GAPI is not initialized")
 			return
 		}
 		pChan := puppetChan()
@@ -100,7 +100,12 @@ func (obj *GAPI) SwitchStream() chan error {
 				}
 				log.Printf("Puppet: Generating new graph...")
 				pChan = puppetChan() // TODO: okay to update interval in case it changed?
-				ch <- nil            // trigger a run
+				select {
+				case ch <- nil: // trigger a run (send a msg)
+				// unblock if we exit while waiting to send!
+				case <-obj.closeChan:
+					return
+				}
 			case <-obj.closeChan:
 				return
 			}
@@ -112,7 +117,7 @@ func (obj *GAPI) SwitchStream() chan error {
 // Close shuts down the Puppet GAPI.
 func (obj *GAPI) Close() error {
 	if !obj.initialized {
-		return fmt.Errorf("Puppet: GAPI is not initialized!")
+		return fmt.Errorf("the puppet GAPI is not initialized")
 	}
 	close(obj.closeChan)
 	obj.wg.Wait()

puppet/puppet.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -32,7 +32,8 @@ import (
 const (
 	// PuppetYAMLBufferSize is the maximum buffer size for the yaml input data
 	PuppetYAMLBufferSize = 65535
-	Debug                = false // FIXME: integrate with global debug flag
+	// Debug is a local debug constant used in this module
+	Debug = false // FIXME: integrate with global debug flag
 )
 
 func runPuppetCommand(cmd *exec.Cmd) ([]byte, error) {
@@ -115,8 +116,8 @@ func ParseConfigFromPuppet(puppetParam, puppetConf string) *yamlgraph.GraphConfi
 	return &config
 }
 
-// PuppetInterval returns the graph refresh interval from the puppet configuration.
-func PuppetInterval(puppetConf string) int {
+// RefreshInterval returns the graph refresh interval from the puppet configuration.
+func RefreshInterval(puppetConf string) int {
 	if Debug {
 		log.Printf("Puppet: determining graph refresh interval")
 	}

recwatch/configwatch.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -59,15 +59,23 @@ func (obj *ConfigWatcher) Add(file ...string) {
 		ch := obj.ConfigWatch(file[0])
 		for {
 			select {
-			case e := <-ch:
+			case e, ok := <-ch:
+				if !ok { // channel closed
+					return
+				}
 				if e != nil {
 					obj.errorchan <- e
 					return
 				}
-				obj.ch <- file[0]
+				select {
+				case obj.ch <- file[0]: // send on channel
+				case <-obj.closechan:
+					return // never mind, close early!
+				}
 				continue
-			case <-obj.closechan:
-				return
+				// not needed, closes via ConfigWatch() chan close
+				//case <-obj.closechan:
+				//	return
 			}
 		}
 	}()
@@ -126,7 +134,12 @@ func (obj *ConfigWatcher) ConfigWatch(file string) chan error {
 					close(ch)
 					return
 				}
-				ch <- nil // send event!
+				select {
+				case ch <- nil: // send event!
+				case <-obj.closechan:
+					close(ch)
+					return
+				}
 			}
 		}
 		//close(ch)

recwatch/flags.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify

recwatch/recwatch.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -51,10 +51,10 @@ type RecWatcher struct {
 	watcher  *fsnotify.Watcher
 	watches  map[string]struct{}
 	events   chan Event // one channel for events and err...
-	once     sync.Once
+	closed   bool       // is the events channel closed?
+	mutex    sync.Mutex // lock guarding the channel closing
 	wg       sync.WaitGroup
 	exit     chan struct{}
-	closeErr error
 }
 
 // NewRecWatcher creates an initializes a new recursive watcher.
@@ -87,11 +87,22 @@ func (obj *RecWatcher) Init() error {
 		}
 	}
 
+	obj.wg.Add(1)
 	go func() {
+		defer obj.wg.Done()
 		if err := obj.Watch(); err != nil {
-			obj.events <- Event{Error: err}
+			// we need this mutex, because if we Init and then Close
+			// immediately, this can send after closed which panics!
+			obj.mutex.Lock()
+			if !obj.closed {
+				select {
+				case obj.events <- Event{Error: err}:
+				case <-obj.exit:
+					// pass
+				}
+			}
+			obj.mutex.Unlock()
 		}
-		obj.Close()
 	}()
 	return nil
 }
@@ -106,26 +117,18 @@ func (obj *RecWatcher) Init() error {
 
 // Close shuts down the watcher.
 func (obj *RecWatcher) Close() error {
-	obj.once.Do(obj.close) // don't cause the channel to close twice
-	return obj.closeErr
-}
-
-// This close function is the function that actually does the close work. Don't
-// call it more than once!
-func (obj *RecWatcher) close() {
 	var err error
 	close(obj.exit) // send exit signal
 	obj.wg.Wait()
 	if obj.watcher != nil {
 		err = obj.watcher.Close()
 		obj.watcher = nil
-		// TODO: should we send the close error?
-		//if err != nil {
-		//	obj.events <- Event{Error: err}
-		//}
 	}
+	obj.mutex.Lock() // FIXME: I don't think this mutex is needed anymore...
+	obj.closed = true
 	close(obj.events)
-	obj.closeErr = err // set the error
+	obj.mutex.Unlock()
+	return err
 }
 
 // Events returns a channel of events. These include events for errors.
@@ -134,10 +137,8 @@ func (obj *RecWatcher) Events() chan Event { return obj.events }
 // Watch is the primary listener for this resource and it outputs events.
 func (obj *RecWatcher) Watch() error {
 	if obj.watcher == nil {
-		return fmt.Errorf("Watcher is not initialized!")
+		return fmt.Errorf("the watcher is not initialized")
 	}
-	obj.wg.Add(1)
-	defer obj.wg.Done()
 
 	patharray := util.PathSplit(obj.safename) // tokenize the path
 	var index = len(patharray)                // starting index
@@ -169,11 +170,11 @@ func (obj *RecWatcher) Watch() error {
 				// no space left on device, out of inotify watches
 				// TODO: consider letting the user fall back to
 				// polling if they hit this error very often...
-				return fmt.Errorf("Out of inotify watches: %v", err)
+				return fmt.Errorf("out of inotify watches: %v", err)
 			} else if os.IsPermission(err) {
-				return fmt.Errorf("Permission denied adding a watch: %v", err)
+				return fmt.Errorf("permission denied adding a watch: %v", err)
 			}
-			return fmt.Errorf("Unknown error: %v", err)
+			return fmt.Errorf("unknown error: %v", err)
 		}
 
 		select {
@@ -236,6 +237,13 @@ func (obj *RecWatcher) Watch() error {
 					index--
 				}
 
+				// when the file is moved, remove the watcher and add a new one,
+				// so we stop tracking the old inode.
+				if deltaDepth >= 0 && (event.Op&fsnotify.Rename == fsnotify.Rename) {
+					obj.watcher.Remove(current)
+					obj.watcher.Add(current)
+				}
+
 				// we must be a parent watcher, so descend in
 				if deltaDepth < 0 {
 					// XXX: we can block here due to: https://github.com/fsnotify/fsnotify/issues/123
@@ -272,11 +280,16 @@ func (obj *RecWatcher) Watch() error {
 			if send {
 				send = false
 				// only invalid state on certain types of events
-				obj.events <- Event{Error: nil, Body: &event}
+				select {
+				// exit even when we're blocked on event sending
+				case obj.events <- Event{Error: nil, Body: &event}:
+				case <-obj.exit:
+					return fmt.Errorf("pending event not sent")
+				}
 			}
 
 		case err := <-obj.watcher.Errors:
-			return fmt.Errorf("Unknown watcher error: %v", err)
+			return fmt.Errorf("unknown watcher error: %v", err)
 
 		case <-obj.exit:
 			return nil

remote/remote.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -63,6 +63,7 @@ import (
 
 	cv "github.com/purpleidea/mgmt/converger"
 	"github.com/purpleidea/mgmt/util"
+	"github.com/purpleidea/mgmt/util/semaphore"
 	"github.com/purpleidea/mgmt/yamlgraph"
 
 	multierr "github.com/hashicorp/go-multierror"
@@ -156,11 +157,11 @@ func (obj *SSH) Sftp() error {
 	var err error
 
 	if obj.client == nil {
-		return fmt.Errorf("Not dialed!")
+		return fmt.Errorf("not dialed")
 	}
 	// this check is needed because the golang path.Base function is weird!
 	if strings.HasSuffix(obj.file, "/") {
-		return fmt.Errorf("File must not be a directory.")
+		return fmt.Errorf("file must not be a directory")
 	}
 
 	// we run local operations first so that remote clean up is easier...
@@ -254,7 +255,7 @@ func (obj *SSH) Sftp() error {
 	// make file executable; don't cache this in case it didn't ever happen
 	// TODO: do we want the group or other bits set?
 	if err := obj.sftp.Chmod(obj.execpath, 0770); err != nil {
-		return fmt.Errorf("Can't set file mode bits!")
+		return fmt.Errorf("can't set file mode bits")
 	}
 
 	// copy graph file
@@ -273,7 +274,7 @@ func (obj *SSH) Sftp() error {
 // SftpGraphCopy is a helper function used for re-copying the graph definition.
 func (obj *SSH) SftpGraphCopy() (int64, error) {
 	if obj.filepath == "" {
-		return -1, fmt.Errorf("Sftp session isn't ready yet!")
+		return -1, fmt.Errorf("sftp session isn't ready yet")
 	}
 	return obj.SftpCopy(obj.file, obj.filepath)
 }
@@ -281,7 +282,7 @@ func (obj *SSH) SftpGraphCopy() (int64, error) {
 // SftpCopy is a simple helper function that runs a local -> remote sftp copy.
 func (obj *SSH) SftpCopy(src, dst string) (int64, error) {
 	if obj.sftp == nil {
-		return -1, fmt.Errorf("Sftp session is not active!")
+		return -1, fmt.Errorf("sftp session is not active")
 	}
 	var err error
 	// TODO: add a check to make sure we don't run two copies of this
@@ -313,7 +314,7 @@ func (obj *SSH) SftpCopy(src, dst string) (int64, error) {
 		return n, fmt.Errorf("Can't copy to remote path: %v", err)
 	}
 	if n <= 0 {
-		return n, fmt.Errorf("Zero bytes copied!")
+		return n, fmt.Errorf("zero bytes copied")
 	}
 	return n, nil
 }
@@ -391,10 +392,10 @@ func (obj *SSH) Tunnel() error {
 	var err error
 
 	if len(obj.clientURLs) < 1 {
-		return fmt.Errorf("Need at least one client URL to tunnel!")
+		return fmt.Errorf("need at least one client URL to tunnel")
 	}
 	if len(obj.remoteURLs) < 1 {
-		return fmt.Errorf("Need at least one remote URL to tunnel!")
+		return fmt.Errorf("need at least one remote URL to tunnel")
 	}
 
 	// TODO: do something less arbitrary about which one we pick?
@@ -477,10 +478,10 @@ func (obj *SSH) TunnelClose() error {
 // Exec runs the binary on the remote server.
 func (obj *SSH) Exec() error {
 	if obj.execpath == "" {
-		return fmt.Errorf("Must have a binary path to execute!")
+		return fmt.Errorf("must have a binary path to execute")
 	}
 	if obj.filepath == "" {
-		return fmt.Errorf("Must have a graph definition to run!")
+		return fmt.Errorf("must have a graph definition to run")
 	}
 
 	var err error
@@ -698,16 +699,16 @@ type Remotes struct {
 	converger    cv.Converger
 	convergerCb  func(func(map[string]bool) error) (func(), error)
 
-	wg                 sync.WaitGroup             // keep track of each running SSH connection
-	lock               sync.Mutex                 // mutex for access to sshmap
-	sshmap             map[string]*SSH            // map to each SSH struct with the remote as the key
-	exiting            bool                       // flag to let us know if we're exiting
-	exitChan           chan struct{}              // closes when we should exit
-	semaphore          Semaphore                  // counting semaphore to limit concurrent connections
-	hostnames          []string                   // list of hostnames we've seen so far
-	cuid               cv.ConvergerUID            // convergerUID for the remote itself
-	cuids              map[string]cv.ConvergerUID // map to each SSH struct with the remote as the key
-	callbackCancelFunc func()                     // stored callback function cancel function
+	wg                 sync.WaitGroup       // keep track of each running SSH connection
+	lock               sync.Mutex           // mutex for access to sshmap
+	sshmap             map[string]*SSH      // map to each SSH struct with the remote as the key
+	exiting            bool                 // flag to let us know if we're exiting
+	exitChan           chan struct{}        // closes when we should exit
+	semaphore          *semaphore.Semaphore // counting semaphore to limit concurrent connections
+	hostnames          []string             // list of hostnames we've seen so far
+	cuid               cv.UID               // convergerUID for the remote itself
+	cuids              map[string]cv.UID    // map to each SSH struct with the remote as the key
+	callbackCancelFunc func()               // stored callback function cancel function
 
 	flags Flags // constant runtime values
 }
@@ -730,9 +731,9 @@ func NewRemotes(clientURLs, remoteURLs []string, noop bool, remotes []string, fi
 		convergerCb:  convergerCb,
 		sshmap:       make(map[string]*SSH),
 		exitChan:     make(chan struct{}),
-		semaphore:    NewSemaphore(int(cConns)),
+		semaphore:    semaphore.NewSemaphore(int(cConns)),
 		hostnames:    make([]string, len(remotes)),
-		cuids:        make(map[string]cv.ConvergerUID),
+		cuids:        make(map[string]cv.UID),
 		flags:        flags,
 	}
 }
@@ -772,7 +773,7 @@ func (obj *Remotes) NewSSH(file string) (*SSH, error) {
 	}
 	host = x[0]
 	if host == "" {
-		return nil, fmt.Errorf("Empty hostname!")
+		return nil, fmt.Errorf("empty hostname")
 	}
 
 	user := defaultUser // default
@@ -795,7 +796,7 @@ func (obj *Remotes) NewSSH(file string) (*SSH, error) {
 	}
 
 	if len(auth) == 0 {
-		return nil, fmt.Errorf("No authentication methods available!")
+		return nil, fmt.Errorf("no authentication methods available")
 	}
 
 	//hostname := config.Hostname // TODO: optionally specify local hostname somehow
@@ -804,7 +805,7 @@ func (obj *Remotes) NewSSH(file string) (*SSH, error) {
 		hostname = host // default to above
 	}
 	if util.StrInList(hostname, obj.hostnames) {
-		return nil, fmt.Errorf("Remote: Hostname `%s` already exists!", hostname)
+		return nil, fmt.Errorf("Remote: Hostname `%s` already exists", hostname)
 	}
 	obj.hostnames = append(obj.hostnames, hostname)
 
@@ -830,7 +831,7 @@ func (obj *Remotes) NewSSH(file string) (*SSH, error) {
 // sshKeyAuth is a helper function to get the ssh key auth struct needed
 func (obj *Remotes) sshKeyAuth() (ssh.AuthMethod, error) {
 	if obj.sshPrivIdRsa == "" {
-		return nil, fmt.Errorf("Empty path specified!")
+		return nil, fmt.Errorf("empty path specified")
 	}
 	p := ""
 	// TODO: this doesn't match strings of the form: ~james/.ssh/id_rsa
@@ -843,7 +844,7 @@ func (obj *Remotes) sshKeyAuth() (ssh.AuthMethod, error) {
 		p = path.Join(usr.HomeDir, obj.sshPrivIdRsa[len("~/"):])
 	}
 	if p == "" {
-		return nil, fmt.Errorf("Empty path specified!")
+		return nil, fmt.Errorf("empty path specified")
 	}
 	// A public key may be used to authenticate against the server by using
 	// an unencrypted PEM-encoded private key file. If you have an encrypted
@@ -892,7 +893,7 @@ func (obj *Remotes) passwordCallback(user, host string) func() (string, error) {
 		case e := <-failchan:
 			return "", e
 		case <-util.TimeAfterOrBlock(timeout):
-			return "", fmt.Errorf("Interactive timeout reached!")
+			return "", fmt.Errorf("interactive timeout reached")
 		}
 	}
 	return cb
@@ -1078,29 +1079,6 @@ func cleanURL(s string) string {
 	return u.Host
 }
 
-// Semaphore is a counting semaphore.
-type Semaphore chan struct{}
-
-// NewSemaphore creates a new semaphore.
-func NewSemaphore(size int) Semaphore {
-	return make(Semaphore, size)
-}
-
-// P acquires n resources.
-func (s Semaphore) P(n int) {
-	e := struct{}{}
-	for i := 0; i < n; i++ {
-		s <- e // acquire one
-	}
-}
-
-// V releases n resources.
-func (s Semaphore) V(n int) {
-	for i := 0; i < n; i++ {
-		<-s // release one
-	}
-}
-
 // combinedWriter mimics what the ssh.CombinedOutput command does.
 type combinedWriter struct {
 	b  bytes.Buffer

resources/augeas.go

@@ -0,0 +1,305 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// +build !noaugeas
+
+package resources
+
+import (
+	"encoding/gob"
+	"fmt"
+	"log"
+	"os"
+	"strings"
+
+	"github.com/purpleidea/mgmt/recwatch"
+
+	errwrap "github.com/pkg/errors"
+	// FIXME: we vendor go/augeas because master requires augeas 1.6.0
+	// and libaugeas-dev-1.6.0 is not yet available in a PPA.
+	"honnef.co/go/augeas"
+)
+
+const (
+	// NS is a namespace for augeas operations
+	NS = "Xmgmt"
+)
+
+func init() {
+	gob.Register(&AugeasRes{})
+}
+
+// AugeasRes is a resource that enables you to use the augeas resource.
+// Currently only allows you to change simple files (e.g sshd_config).
+type AugeasRes struct {
+	BaseRes `yaml:",inline"`
+
+	// File is the path to the file targeted by this resource.
+	File string `yaml:"file"`
+
+	// Lens is the lens used by this resource. If specified, mgmt
+	// will lower the augeas overhead by only loading that lens.
+	Lens string `yaml:"lens"`
+
+	// Sets is a list of changes that will be applied to the file, in the form of
+	// ["path", "value"]. mgmt will run augeas.Get() before augeas.Set(), to
+	// prevent changing the file when it is not needed.
+	Sets []AugeasSet `yaml:"sets"`
+
+	recWatcher *recwatch.RecWatcher // used to watch the changed files
+}
+
+// AugeasSet represents a key/value pair of settings to be applied.
+type AugeasSet struct {
+	Path  string `yaml:"path"`  // The relative path to the value to be changed.
+	Value string `yaml:"value"` // The value to be set on the given Path.
+}
+
+// Default returns some sensible defaults for this resource.
+func (obj *AugeasRes) Default() Res {
+	return &AugeasRes{
+		BaseRes: BaseRes{
+			MetaParams: DefaultMetaParams, // force a default
+		},
+	}
+}
+
+// Validate if the params passed in are valid data.
+func (obj *AugeasRes) Validate() error {
+	if !strings.HasPrefix(obj.File, "/") {
+		return fmt.Errorf("the File param should start with a slash")
+	}
+	if obj.Lens != "" && !strings.HasSuffix(obj.Lens, ".lns") {
+		return fmt.Errorf("the Lens param should have a .lns suffix")
+	}
+	if (obj.Lens == "") != (obj.File == "") {
+		return fmt.Errorf("the File and Lens params must be specified together")
+	}
+	return obj.BaseRes.Validate()
+}
+
+// Init initiates the resource.
+func (obj *AugeasRes) Init() error {
+	obj.BaseRes.kind = "augeas"
+	return obj.BaseRes.Init() // call base init, b/c we're overriding
+}
+
+// Watch is the primary listener for this resource and it outputs events.
+// Taken from the File resource.
+// FIXME: DRY - This is taken from the file resource
+func (obj *AugeasRes) Watch() error {
+	var err error
+	obj.recWatcher, err = recwatch.NewRecWatcher(obj.File, false)
+	if err != nil {
+		return err
+	}
+	defer obj.recWatcher.Close()
+
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
+	var send = false // send event?
+	var exit *error
+
+	for {
+		if obj.debug {
+			log.Printf("%s[%s]: Watching: %s", obj.Kind(), obj.GetName(), obj.File) // attempting to watch...
+		}
+
+		select {
+		case event, ok := <-obj.recWatcher.Events():
+			if !ok { // channel shutdown
+				return nil
+			}
+			if err := event.Error; err != nil {
+				return errwrap.Wrapf(err, "Unknown %s[%s] watcher error", obj.Kind(), obj.GetName())
+			}
+			if obj.debug { // don't access event.Body if event.Error isn't nil
+				log.Printf("%s[%s]: Event(%s): %v", obj.Kind(), obj.GetName(), event.Body.Name, event.Body.Op)
+			}
+			send = true
+			obj.StateOK(false) // dirty
+
+		case event := <-obj.Events():
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
+			}
+			//obj.StateOK(false) // dirty // these events don't invalidate state
+		}
+
+		// do all our event sending all together to avoid duplicate msgs
+		if send {
+			send = false
+			obj.Event()
+		}
+	}
+}
+
+// checkApplySet runs CheckApply for one element of the AugeasRes.Set
+func (obj *AugeasRes) checkApplySet(apply bool, ag *augeas.Augeas, set AugeasSet) (bool, error) {
+	fullpath := fmt.Sprintf("/files/%v/%v", obj.File, set.Path)
+
+	// We do not check for errors because errors are also thrown when
+	// the path does not exist.
+	if getValue, _ := ag.Get(fullpath); set.Value == getValue {
+		// The value is what we expect, return directly
+		return true, nil
+	}
+
+	if !apply {
+		// If noop, we can return here directly. We return with
+		// nil even if err is not nil because it does not mean
+		// there is an error.
+		return false, nil
+	}
+
+	if err := ag.Set(fullpath, set.Value); err != nil {
+		return false, errwrap.Wrapf(err, "augeas: error while setting value")
+	}
+
+	return false, nil
+}
+
+// CheckApply method for Augeas resource.
+func (obj *AugeasRes) CheckApply(apply bool) (bool, error) {
+	log.Printf("%s[%s]: CheckApply: %s", obj.Kind(), obj.GetName(), obj.File)
+	// By default we do not set any option to augeas, we use the defaults.
+	opts := augeas.None
+	if obj.Lens != "" {
+		// if the lens is specified, we can speed up augeas by not
+		// loading everything. Without this option, augeas will try to
+		// read all the files it knows in the complete filesystem.
+		// e.g. to change /etc/ssh/sshd_config, it would load /etc/hosts, /etc/ntpd.conf, etc...
+		opts = augeas.NoModlAutoload
+	}
+
+	// Initiate augeas
+	ag, err := augeas.New("/", "", opts)
+	if err != nil {
+		return false, errwrap.Wrapf(err, "augeas: error while initializing")
+	}
+	defer ag.Close()
+
+	if obj.Lens != "" {
+		// If the lens is set, load the lens for the file we want to edit.
+		// We pick Xmgmt, as this name will not collide with any other lens name.
+		// We do not pick Mgmt as in the future there might be an Mgmt lens.
+		// https://github.com/hercules-team/augeas/wiki/Loading-specific-files
+		if err = ag.Set(fmt.Sprintf("/augeas/load/%s/lens", NS), obj.Lens); err != nil {
+			return false, errwrap.Wrapf(err, "augeas: error while initializing lens")
+		}
+		if err = ag.Set(fmt.Sprintf("/augeas/load/%s/incl", NS), obj.File); err != nil {
+			return false, errwrap.Wrapf(err, "augeas: error while initializing incl")
+		}
+		if err = ag.Load(); err != nil {
+			return false, errwrap.Wrapf(err, "augeas: error while loading")
+		}
+	}
+
+	checkOK := true
+	for _, set := range obj.Sets {
+		if setCheckOK, err := obj.checkApplySet(apply, &ag, set); err != nil {
+			return false, errwrap.Wrapf(err, "augeas: error during CheckApply of one Set")
+		} else if !setCheckOK {
+			checkOK = false
+		}
+	}
+
+	// If the state is correct or we can't apply, return early.
+	if checkOK || !apply {
+		return checkOK, nil
+	}
+
+	log.Printf("%s[%s]: changes needed, saving", obj.Kind(), obj.GetName())
+	if err = ag.Save(); err != nil {
+		return false, errwrap.Wrapf(err, "augeas: error while saving augeas values")
+	}
+
+	// FIXME: Workaround for https://github.com/dominikh/go-augeas/issues/13
+	// To be fixed upstream.
+	if obj.File != "" {
+		if _, err := os.Stat(obj.File); os.IsNotExist(err) {
+			return false, errwrap.Wrapf(err, "augeas: error: file does not exist")
+		}
+	}
+
+	return false, nil
+}
+
+// AugeasUID is the UID struct for AugeasRes.
+type AugeasUID struct {
+	BaseUID
+	name string
+}
+
+// AutoEdges returns the AutoEdge interface. In this case no autoedges are used.
+func (obj *AugeasRes) AutoEdges() AutoEdge {
+	return nil
+}
+
+// UIDs includes all params to make a unique identification of this object.
+func (obj *AugeasRes) UIDs() []ResUID {
+	x := &AugeasUID{
+		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
+		name:    obj.Name,
+	}
+	return []ResUID{x}
+}
+
+// GroupCmp returns whether two resources can be grouped together or not.
+func (obj *AugeasRes) GroupCmp(r Res) bool {
+	return false // Augeas commands can not be grouped together.
+}
+
+// Compare two resources and return if they are equivalent.
+func (obj *AugeasRes) Compare(res Res) bool {
+	switch res.(type) {
+	// we can only compare AugeasRes to others of the same resource
+	case *AugeasRes:
+		res := res.(*AugeasRes)
+		if !obj.BaseRes.Compare(res) { // call base Compare
+			return false
+		}
+		if obj.Name != res.Name {
+			return false
+		}
+	default:
+		return false
+	}
+	return true
+}
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *AugeasRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes AugeasRes // indirection to avoid infinite recursion
+
+	def := obj.Default()        // get the default
+	res, ok := def.(*AugeasRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to AugeasRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = AugeasRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/augeas_disabled.go

@@ -0,0 +1,24 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// +build noaugeas
+
+package resources
+
+// AugeasRes represents the fields of the Augeas resource. Since this file is
+// only invoked with the tag "noaugeas", we do not need any fields here.
+type AugeasRes struct {
+}

resources/exec.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -25,9 +25,8 @@ import (
 	"log"
 	"os/exec"
 	"strings"
-	"time"
+	"syscall"
 
-	"github.com/purpleidea/mgmt/event"
 	"github.com/purpleidea/mgmt/util"
 
 	errwrap "github.com/pkg/errors"
@@ -40,7 +39,6 @@ func init() {
 // ExecRes is an exec resource for running commands.
 type ExecRes struct {
 	BaseRes    `yaml:",inline"`
-	State      string `yaml:"state"`      // state: exists/present?, absent, (undefined?)
 	Cmd        string `yaml:"cmd"`        // the command to run
 	Shell      string `yaml:"shell"`      // the (optional) shell to use to run the cmd
 	Timeout    int    `yaml:"timeout"`    // the cmd timeout in seconds
@@ -48,49 +46,32 @@ type ExecRes struct {
 	WatchShell string `yaml:"watchshell"` // the (optional) shell to use to run the watch cmd
 	IfCmd      string `yaml:"ifcmd"`      // the if command to run
 	IfShell    string `yaml:"ifshell"`    // the (optional) shell to use to run the if cmd
-	PollInt    int    `yaml:"pollint"`    // the poll interval for the ifcmd
 }
 
-// NewExecRes is a constructor for this resource. It also calls Init() for you.
-func NewExecRes(name, cmd, shell string, timeout int, watchcmd, watchshell, ifcmd, ifshell string, pollint int, state string) (*ExecRes, error) {
-	obj := &ExecRes{
+// Default returns some sensible defaults for this resource.
+func (obj *ExecRes) Default() Res {
+	return &ExecRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		Cmd:        cmd,
-		Shell:      shell,
-		Timeout:    timeout,
-		WatchCmd:   watchcmd,
-		WatchShell: watchshell,
-		IfCmd:      ifcmd,
-		IfShell:    ifshell,
-		PollInt:    pollint,
-		State:      state,
 	}
-	return obj, obj.Init()
+}
+
+// Validate if the params passed in are valid data.
+func (obj *ExecRes) Validate() error {
+	if obj.Cmd == "" { // this is the only thing that is really required
+		return fmt.Errorf("command can't be empty")
+	}
+
+	return obj.BaseRes.Validate()
 }
 
 // Init runs some startup code for this resource.
 func (obj *ExecRes) Init() error {
-	obj.BaseRes.kind = "Exec"
+	obj.BaseRes.kind = "exec"
 	return obj.BaseRes.Init() // call base init, b/c we're overriding
 }
 
-// Validate if the params passed in are valid data.
-// FIXME: where should this get called ?
-func (obj *ExecRes) Validate() error {
-	if obj.Cmd == "" { // this is the only thing that is really required
-		return fmt.Errorf("Command can't be empty!")
-	}
-
-	// if we have a watch command, then we don't poll with the if command!
-	if obj.WatchCmd != "" && obj.PollInt > 0 {
-		return fmt.Errorf("Don't poll when we have a watch command.")
-	}
-
-	return nil
-}
-
 // BufioChanScanner wraps the scanner output in a channel.
 func (obj *ExecRes) BufioChanScanner(scanner *bufio.Scanner) (chan string, chan error) {
 	ch, errch := make(chan string), make(chan error)
@@ -110,26 +91,9 @@ func (obj *ExecRes) BufioChanScanner(scanner *bufio.Scanner) (chan string, chan
 }
 
 // Watch is the primary listener for this resource and it outputs events.
-func (obj *ExecRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
-	}
-
+func (obj *ExecRes) Watch() error {
 	var send = false // send event?
-	var exit = false
+	var exit *error
 	bufioch, errch := make(chan string), make(chan error)
 
 	if obj.WatchCmd != "" {
@@ -149,71 +113,65 @@ func (obj *ExecRes) Watch(processChan chan event.Event) error {
 		}
 		cmd := exec.Command(cmdName, cmdArgs...)
 		//cmd.Dir = "" // look for program in pwd ?
+		// ignore signals sent to parent process (we're in our own group)
+		cmd.SysProcAttr = &syscall.SysProcAttr{
+			Setpgid: true,
+			Pgid:    0,
+		}
 
 		cmdReader, err := cmd.StdoutPipe()
 		if err != nil {
-			return errwrap.Wrapf(err, "Error creating StdoutPipe for Cmd")
+			return errwrap.Wrapf(err, "error creating StdoutPipe for Cmd")
 		}
 		scanner := bufio.NewScanner(cmdReader)
 
-		defer cmd.Wait() // XXX: is this necessary?
+		defer cmd.Wait() // wait for the command to exit before return!
 		defer func() {
 			// FIXME: without wrapping this in this func it panic's
-			// when running examples/graph8d.yaml
-			cmd.Process.Kill() // TODO: is this necessary?
+			// when running certain graphs... why?
+			cmd.Process.Kill() // shutdown the Watch command on exit
 		}()
 		if err := cmd.Start(); err != nil {
-			return errwrap.Wrapf(err, "Error starting Cmd")
+			return errwrap.Wrapf(err, "error starting Cmd")
 		}
 
 		bufioch, errch = obj.BufioChanScanner(scanner)
 	}
 
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
 	for {
-		obj.SetState(ResStateWatching) // reset
 		select {
 		case text := <-bufioch:
-			cuid.SetConverged(false)
 			// each time we get a line of output, we loop!
 			log.Printf("%s[%s]: Watch output: %s", obj.Kind(), obj.GetName(), text)
 			if text != "" {
 				send = true
+				obj.StateOK(false) // something made state dirty
 			}
 
 		case err := <-errch:
-			cuid.SetConverged(false)
 			if err == nil { // EOF
 				// FIXME: add an "if watch command ends/crashes"
 				// restart or generate error option
-				return fmt.Errorf("Reached EOF")
+				return fmt.Errorf("reached EOF")
 			}
 			// error reading input?
-			return errwrap.Wrapf(err, "Unknown error")
+			return errwrap.Wrapf(err, "unknown error")
 
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
-			if exit, send = obj.ReadEvent(&event); exit {
-				return nil // exit
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
-
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true) // converged!
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
-			startup = true // startup finished
 			send = false
-			// it is okay to invalidate the clean state on poke too
-			obj.StateOK(false) // something made state dirty
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -221,24 +179,12 @@ func (obj *ExecRes) Watch(processChan chan event.Event) error {
 // CheckApply checks the resource state and applies the resource if the bool
 // input is true. It returns error info and if the state check passed or not.
 // TODO: expand the IfCmd to be a list of commands
-func (obj *ExecRes) CheckApply(apply bool) (checkOK bool, err error) {
+func (obj *ExecRes) CheckApply(apply bool) (bool, error) {
+	// If we receive a refresh signal, then the engine skips the IsStateOK()
+	// check and this will run. It is still guarded by the IfCmd, but it can
+	// have a chance to execute, and all without the check of obj.Refresh()!
 
-	// if there is a watch command, but no if command, run based on state
-	if obj.WatchCmd != "" && obj.IfCmd == "" {
-		if obj.IsStateOK() { // FIXME: this is done by engine now...
-			return true, nil
-		}
-
-		// if there is no watcher, but there is an onlyif check, run it to see
-	} else if obj.IfCmd != "" { // && obj.WatchCmd == ""
-		// there is a watcher, but there is also an if command
-		//} else if obj.IfCmd != "" && obj.WatchCmd != "" {
-
-		if obj.PollInt > 0 { // && obj.WatchCmd == ""
-			// XXX: have the Watch() command output onlyif poll events...
-			// XXX: we can optimize by saving those results for returning here
-			// return XXX
-		}
+	if obj.IfCmd != "" { // if there is no onlyif check, we should just run
 
 		var cmdName string
 		var cmdArgs []string
@@ -254,18 +200,17 @@ func (obj *ExecRes) CheckApply(apply bool) (checkOK bool, err error) {
 			cmdName = obj.IfShell // usually bash, or sh
 			cmdArgs = []string{"-c", obj.IfCmd}
 		}
-		err = exec.Command(cmdName, cmdArgs...).Run()
-		if err != nil {
+		cmd := exec.Command(cmdName, cmdArgs...)
+		// ignore signals sent to parent process (we're in our own group)
+		cmd.SysProcAttr = &syscall.SysProcAttr{
+			Setpgid: true,
+			Pgid:    0,
+		}
+		if err := cmd.Run(); err != nil {
 			// TODO: check exit value
 			return true, nil // don't run
 		}
 
-		// if there is no watcher and no onlyif check, assume we should run
-	} else { // if obj.WatchCmd == "" && obj.IfCmd == "" {
-		// just run if state is dirty
-		if obj.IsStateOK() { // FIXME: this is done by engine now...
-			return true, nil
-		}
 	}
 
 	// state is not okay, no work done, exit, but without error
@@ -292,11 +237,17 @@ func (obj *ExecRes) CheckApply(apply bool) (checkOK bool, err error) {
 	}
 	cmd := exec.Command(cmdName, cmdArgs...)
 	//cmd.Dir = "" // look for program in pwd ?
+	// ignore signals sent to parent process (we're in our own group)
+	cmd.SysProcAttr = &syscall.SysProcAttr{
+		Setpgid: true,
+		Pgid:    0,
+	}
+
 	var out bytes.Buffer
 	cmd.Stdout = &out
 
 	if err := cmd.Start(); err != nil {
-		return false, errwrap.Wrapf(err, "Error starting Cmd")
+		return false, errwrap.Wrapf(err, "error starting cmd")
 	}
 
 	timeout := obj.Timeout
@@ -306,16 +257,30 @@ func (obj *ExecRes) CheckApply(apply bool) (checkOK bool, err error) {
 	done := make(chan error)
 	go func() { done <- cmd.Wait() }()
 
+	var err error // error returned by cmd
 	select {
-	case err := <-done:
-		if err != nil {
-			e := errwrap.Wrapf(err, "Error waiting for Cmd")
-			return false, e
-		}
+	case e := <-done:
+		err = e // store
 
 	case <-util.TimeAfterOrBlock(timeout):
-		//cmd.Process.Kill() // TODO: is this necessary?
-		return false, fmt.Errorf("Timeout waiting for Cmd!")
+		cmd.Process.Kill() // TODO: check error?
+		return false, fmt.Errorf("timeout for cmd")
+	}
+
+	// process the err result from cmd, we process non-zero exits here too!
+	exitErr, ok := err.(*exec.ExitError) // embeds an os.ProcessState
+	if err != nil && ok {
+		pStateSys := exitErr.Sys() // (*os.ProcessState) Sys
+		wStatus, ok := pStateSys.(syscall.WaitStatus)
+		if !ok {
+			e := errwrap.Wrapf(err, "error running cmd")
+			return false, e
+		}
+		return false, fmt.Errorf("cmd error, exit status: %d", wStatus.ExitStatus())
+
+	} else if err != nil {
+		e := errwrap.Wrapf(err, "general cmd error")
+		return false, e
 	}
 
 	// TODO: if we printed the stdout while the command is running, this
@@ -328,7 +293,6 @@ func (obj *ExecRes) CheckApply(apply bool) (checkOK bool, err error) {
 		log.Printf("%s[%s]: Command output is:", obj.Kind(), obj.GetName())
 		log.Printf(out.String())
 	}
-	// XXX: return based on exit value!!
 
 	// The state tracking is for exec resources that can't "detect" their
 	// state, and assume it's invalid when the Watch() function triggers.
@@ -346,40 +310,6 @@ type ExecUID struct {
 	// TODO: add more elements here
 }
 
-// IFF aka if and only if they are equivalent, return true. If not, false.
-func (obj *ExecUID) IFF(uid ResUID) bool {
-	res, ok := uid.(*ExecUID)
-	if !ok {
-		return false
-	}
-	if obj.Cmd != res.Cmd {
-		return false
-	}
-	// TODO: add more checks here
-	//if obj.Shell != res.Shell {
-	//	return false
-	//}
-	//if obj.Timeout != res.Timeout {
-	//	return false
-	//}
-	//if obj.WatchCmd != res.WatchCmd {
-	//	return false
-	//}
-	//if obj.WatchShell != res.WatchShell {
-	//	return false
-	//}
-	if obj.IfCmd != res.IfCmd {
-		return false
-	}
-	//if obj.PollInt != res.PollInt {
-	//	return false
-	//}
-	//if obj.State != res.State {
-	//	return false
-	//}
-	return true
-}
-
 // AutoEdges returns the AutoEdge interface. In this case no autoedges are used.
 func (obj *ExecRes) AutoEdges() AutoEdge {
 	// TODO: parse as many exec params to look for auto edges, for example
@@ -387,9 +317,9 @@ func (obj *ExecRes) AutoEdges() AutoEdge {
 	return nil
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *ExecRes) GetUIDs() []ResUID {
+func (obj *ExecRes) UIDs() []ResUID {
 	x := &ExecUID{
 		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
 		Cmd:     obj.Cmd,
@@ -438,10 +368,7 @@ func (obj *ExecRes) Compare(res Res) bool {
 		if obj.IfCmd != res.IfCmd {
 			return false
 		}
-		if obj.PollInt != res.PollInt {
-			return false
-		}
-		if obj.State != res.State {
+		if obj.IfShell != res.IfShell {
 			return false
 		}
 	default:
@@ -449,3 +376,23 @@ func (obj *ExecRes) Compare(res Res) bool {
 	}
 	return true
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *ExecRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes ExecRes // indirection to avoid infinite recursion
+
+	def := obj.Default()      // get the default
+	res, ok := def.(*ExecRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to ExecRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = ExecRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/file.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -27,12 +27,13 @@ import (
 	"io/ioutil"
 	"log"
 	"os"
+	"os/user"
 	"path"
 	"path/filepath"
+	"strconv"
 	"strings"
-	"time"
+	"syscall"
 
-	"github.com/purpleidea/mgmt/event"
 	"github.com/purpleidea/mgmt/recwatch"
 	"github.com/purpleidea/mgmt/util"
 
@@ -52,6 +53,9 @@ type FileRes struct {
 	Content    *string `yaml:"content"` // nil to mark as undefined
 	Source     string  `yaml:"source"`  // file path for source content
 	State      string  `yaml:"state"`   // state: exists/present?, absent, (undefined?)
+	Owner      string  `yaml:"owner"`
+	Group      string  `yaml:"group"`
+	Mode       string  `yaml:"mode"`
 	Recurse    bool    `yaml:"recurse"`
 	Force      bool    `yaml:"force"`
 	path       string  // computed path
@@ -60,44 +64,105 @@ type FileRes struct {
 	recWatcher *recwatch.RecWatcher
 }
 
-// NewFileRes is a constructor for this resource. It also calls Init() for you.
-func NewFileRes(name, path, dirname, basename string, content *string, source, state string, recurse, force bool) (*FileRes, error) {
-	obj := &FileRes{
+// Default returns some sensible defaults for this resource.
+func (obj *FileRes) Default() Res {
+	return &FileRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		Path:     path,
-		Dirname:  dirname,
-		Basename: basename,
-		Content:  content,
-		Source:   source,
-		State:    state,
-		Recurse:  recurse,
-		Force:    force,
+		State: "exists",
 	}
-	return obj, obj.Init()
+}
+
+// Validate reports any problems with the struct definition.
+func (obj *FileRes) Validate() error {
+	if obj.Dirname != "" && !strings.HasSuffix(obj.Dirname, "/") {
+		return fmt.Errorf("dirname must end with a slash")
+	}
+
+	if strings.HasPrefix(obj.Basename, "/") {
+		return fmt.Errorf("basename must not start with a slash")
+	}
+
+	if obj.Content != nil && obj.Source != "" {
+		return fmt.Errorf("can't specify both Content and Source")
+	}
+
+	if obj.isDir && obj.Content != nil { // makes no sense
+		return fmt.Errorf("can't specify Content when creating a Dir")
+	}
+
+	if obj.Mode != "" {
+		if _, err := obj.mode(); err != nil {
+			return err
+		}
+	}
+
+	if _, err := obj.uid(); obj.Owner != "" && err != nil {
+		return err
+	}
+
+	if _, err := obj.gid(); obj.Group != "" && err != nil {
+		return err
+	}
+
+	// XXX: should this specify that we create an empty directory instead?
+	//if obj.Source == "" && obj.isDir {
+	//	return fmt.Errorf("Can't specify an empty source when creating a Dir.")
+	//}
+
+	return obj.BaseRes.Validate()
+}
+
+// mode returns the file permission specified on the graph. It doesn't handle
+// the case where the mode is not specified. The caller should check obj.Mode is
+// not empty.
+func (obj *FileRes) mode() (os.FileMode, error) {
+	m, err := strconv.ParseInt(obj.Mode, 8, 32)
+	if err != nil {
+		return os.FileMode(0), errwrap.Wrapf(err, "Mode should be an octal number (%s)", obj.Mode)
+	}
+	return os.FileMode(m), nil
+}
+
+// uid returns the user id for the owner specified in the yaml file graph.
+// Caller should first check obj.Owner is not empty
+func (obj *FileRes) uid() (int, error) {
+	u2, err2 := user.LookupId(obj.Owner)
+	if err2 == nil {
+		return strconv.Atoi(u2.Uid)
+	}
+
+	u, err := user.Lookup(obj.Owner)
+	if err == nil {
+		return strconv.Atoi(u.Uid)
+	}
+
+	return -1, errwrap.Wrapf(err, "owner lookup error (%s)", obj.Owner)
 }
 
 // Init runs some startup code for this resource.
 func (obj *FileRes) Init() error {
 	obj.sha256sum = ""
-	if obj.Path == "" { // use the name as the path default if missing
-		obj.Path = obj.BaseRes.Name
-	}
 	obj.path = obj.GetPath()                     // compute once
 	obj.isDir = strings.HasSuffix(obj.path, "/") // dirs have trailing slashes
 
-	obj.BaseRes.kind = "File"
+	obj.BaseRes.kind = "file"
 	return obj.BaseRes.Init() // call base init, b/c we're overriding
 }
 
 // GetPath returns the actual path to use for this resource. It computes this
 // after analysis of the Path, Dirname and Basename values. Dirs end with slash.
 func (obj *FileRes) GetPath() string {
-	d := util.Dirname(obj.Path)
-	b := util.Basename(obj.Path)
+	p := obj.Path
+	if obj.Path == "" { // use the name as the path default if missing
+		p = obj.BaseRes.Name
+	}
+
+	d := util.Dirname(p)
+	b := util.Basename(p)
 	if obj.Dirname == "" && obj.Basename == "" {
-		return obj.Path
+		return p
 	}
 	if obj.Dirname == "" {
 		return d + obj.Basename
@@ -109,80 +174,40 @@ func (obj *FileRes) GetPath() string {
 	return obj.Dirname + obj.Basename
 }
 
-// Validate reports any problems with the struct definition.
-func (obj *FileRes) Validate() error {
-	if obj.Dirname != "" && !strings.HasSuffix(obj.Dirname, "/") {
-		return fmt.Errorf("Dirname must end with a slash.")
-	}
-
-	if strings.HasPrefix(obj.Basename, "/") {
-		return fmt.Errorf("Basename must not start with a slash.")
-	}
-
-	if obj.Content != nil && obj.Source != "" {
-		return fmt.Errorf("Can't specify both Content and Source.")
-	}
-
-	if obj.isDir && obj.Content != nil { // makes no sense
-		return fmt.Errorf("Can't specify Content when creating a Dir.")
-	}
-
-	// XXX: should this specify that we create an empty directory instead?
-	//if obj.Source == "" && obj.isDir {
-	//	return fmt.Errorf("Can't specify an empty source when creating a Dir.")
-	//}
-
-	return nil
-}
-
 // Watch is the primary listener for this resource and it outputs events.
 // This one is a file watcher for files and directories.
 // Modify with caution, it is probably important to write some test cases first!
 // If the Watch returns an error, it means that something has gone wrong, and it
 // must be restarted. On a clean exit it returns nil.
 // FIXME: Also watch the source directory when using obj.Source !!!
-func (obj *FileRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil // TODO: should this be an error?
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
-	}
-
+func (obj *FileRes) Watch() error {
 	var err error
-	obj.recWatcher, err = recwatch.NewRecWatcher(obj.Path, obj.Recurse)
+	obj.recWatcher, err = recwatch.NewRecWatcher(obj.path, obj.Recurse)
 	if err != nil {
 		return err
 	}
 	defer obj.recWatcher.Close()
 
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
 	var send = false // send event?
-	var exit = false
+	var exit *error
 
 	for {
 		if obj.debug {
-			log.Printf("%s[%s]: Watching: %s", obj.Kind(), obj.GetName(), obj.Path) // attempting to watch...
+			log.Printf("%s[%s]: Watching: %s", obj.Kind(), obj.GetName(), obj.path) // attempting to watch...
 		}
 
-		obj.SetState(ResStateWatching) // reset
 		select {
 		case event, ok := <-obj.recWatcher.Events():
 			if !ok { // channel shutdown
 				return nil
 			}
-			cuid.SetConverged(false)
 			if err := event.Error; err != nil {
-				return errwrap.Wrapf(err, "Unknown %s[%s] watcher error", obj.Kind(), obj.GetName())
+				return errwrap.Wrapf(err, "unknown %s[%s] watcher error", obj.Kind(), obj.GetName())
 			}
 			if obj.debug { // don't access event.Body if event.Error isn't nil
 				log.Printf("%s[%s]: Event(%s): %v", obj.Kind(), obj.GetName(), event.Body.Name, event.Body.Op)
@@ -191,29 +216,16 @@ func (obj *FileRes) Watch(processChan chan event.Event) error {
 			obj.StateOK(false) // dirty
 
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
-			if exit, send = obj.ReadEvent(&event); exit {
-				return nil // exit
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
 			//obj.StateOK(false) // dirty // these events don't invalidate state
-
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true) // converged!
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
-			obj.StateOK(false) // dirty
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
-			startup = true // startup finished
 			send = false
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -239,7 +251,7 @@ type FileInfo struct {
 // ReadDir reads a directory path, and returns a list of enhanced FileInfo's.
 func ReadDir(path string) ([]FileInfo, error) {
 	if !strings.HasSuffix(path, "/") { // dirs have trailing slashes
-		return nil, fmt.Errorf("Path must be a directory.")
+		return nil, fmt.Errorf("path must be a directory")
 	}
 	output := []FileInfo{} // my file info
 	fileInfos, err := ioutil.ReadDir(path)
@@ -253,7 +265,7 @@ func ReadDir(path string) ([]FileInfo, error) {
 		abs := path + smartPath(fi)
 		rel, err := filepath.Rel(path, abs) // NOTE: calls Clean()
 		if err != nil {                     // shouldn't happen
-			return nil, errwrap.Wrapf(err, "ReadDir: Unhandled error")
+			return nil, errwrap.Wrapf(err, "unhandled error in ReadDir")
 		}
 		if fi.IsDir() {
 			rel += "/" // add a trailing slash for dirs
@@ -295,7 +307,7 @@ func (obj *FileRes) fileCheckApply(apply bool, src io.ReadSeeker, dst string, sh
 	srcFile, isFile := src.(*os.File)
 	_, isBytes := src.(*bytes.Reader) // supports seeking!
 	if !isFile && !isBytes {
-		return "", false, fmt.Errorf("Can't open src as either file or buffer!")
+		return "", false, fmt.Errorf("can't open src as either file or buffer")
 	}
 
 	var srcStat os.FileInfo
@@ -307,7 +319,7 @@ func (obj *FileRes) fileCheckApply(apply bool, src io.ReadSeeker, dst string, sh
 		}
 		// TODO: deal with symlinks
 		if !srcStat.Mode().IsRegular() { // can't copy non-regular files or dirs
-			return "", false, fmt.Errorf("Non-regular src file: %s (%q)", srcStat.Name(), srcStat.Mode())
+			return "", false, fmt.Errorf("non-regular src file: %s (%q)", srcStat.Name(), srcStat.Mode())
 		}
 	}
 
@@ -331,12 +343,12 @@ func (obj *FileRes) fileCheckApply(apply bool, src io.ReadSeeker, dst string, sh
 			return "", false, nil
 		}
 		if !obj.Force {
-			return "", false, fmt.Errorf("Can't force dir into file: %s", dst)
+			return "", false, fmt.Errorf("can't force dir into file: %s", dst)
 		}
 
 		cleanDst := path.Clean(dst)
 		if cleanDst == "" || cleanDst == "/" {
-			return "", false, fmt.Errorf("Don't want to remove root!") // safety
+			return "", false, fmt.Errorf("don't want to remove root") // safety
 		}
 		// FIXME: respect obj.Recurse here...
 		// there is a dir here, where we want a file...
@@ -348,7 +360,7 @@ func (obj *FileRes) fileCheckApply(apply bool, src io.ReadSeeker, dst string, sh
 
 	} else if err == nil {
 		if !dstStat.Mode().IsRegular() {
-			return "", false, fmt.Errorf("Non-regular dst file: %s (%q)", dstStat.Name(), dstStat.Mode())
+			return "", false, fmt.Errorf("non-regular dst file: %s (%q)", dstStat.Name(), dstStat.Mode())
 		}
 		if isFile && os.SameFile(srcStat, dstStat) { // same inode, we're done!
 			return "", true, nil
@@ -417,6 +429,54 @@ func (obj *FileRes) fileCheckApply(apply bool, src io.ReadSeeker, dst string, sh
 	return sha256sum, false, dstFile.Sync()
 }
 
+// dirCheckApply is the CheckApply operation for an empty directory.
+func (obj *FileRes) dirCheckApply(apply bool) (bool, error) {
+	// check if the path exists and is a directory
+	st, err := os.Stat(obj.path)
+	if err != nil && !os.IsNotExist(err) {
+		return false, errwrap.Wrapf(err, "error checking file resource existence")
+	}
+
+	if err == nil && st.IsDir() {
+		return true, nil // already a directory, nothing to do
+	}
+	if err == nil && !st.IsDir() && !obj.Force {
+		return false, fmt.Errorf("can't force file into dir: %s", obj.path)
+	}
+
+	if !apply {
+		return false, nil
+	}
+
+	// the path exists and is not a directory
+	// delete the file if force is given
+	if err == nil && !st.IsDir() {
+		log.Printf("dirCheckApply: Removing (force): %s", obj.path)
+		if err := os.Remove(obj.path); err != nil {
+			return false, err
+		}
+	}
+
+	// create the empty directory
+	var mode os.FileMode
+	if obj.Mode != "" {
+		mode, err = obj.mode()
+		if err != nil {
+			return false, err
+		}
+	} else {
+		mode = os.ModePerm
+	}
+
+	if obj.Force {
+		// FIXME: respect obj.Recurse here...
+		// TODO: add recurse limit here
+		return false, os.MkdirAll(obj.path, mode)
+	}
+
+	return false, os.Mkdir(obj.path, mode)
+}
+
 // syncCheckApply is the CheckApply operation for a source and destination dir.
 // It is recursive and can create directories directly, and files via the usual
 // fileCheckApply method. It returns checkOK and error as is normally expected.
@@ -425,7 +485,7 @@ func (obj *FileRes) syncCheckApply(apply bool, src, dst string) (bool, error) {
 		log.Printf("syncCheckApply: %s -> %s", src, dst)
 	}
 	if src == "" || dst == "" {
-		return false, fmt.Errorf("The src and dst must not be empty!")
+		return false, fmt.Errorf("the src and dst must not be empty")
 	}
 
 	var checkOK = true
@@ -435,7 +495,7 @@ func (obj *FileRes) syncCheckApply(apply bool, src, dst string) (bool, error) {
 	dstIsDir := strings.HasSuffix(dst, "/")
 
 	if srcIsDir != dstIsDir {
-		return false, fmt.Errorf("The src and dst must be both either files or directories.")
+		return false, fmt.Errorf("the src and dst must be both either files or directories")
 	}
 
 	if !srcIsDir && !dstIsDir {
@@ -488,10 +548,10 @@ func (obj *FileRes) syncCheckApply(apply bool, src, dst string) (bool, error) {
 				if _, ok := smartDst[relPathFile]; ok {
 					absCleanDst := path.Clean(absDst)
 					if !obj.Force {
-						return false, fmt.Errorf("Can't force file into dir: %s", absCleanDst)
+						return false, fmt.Errorf("can't force file into dir: %s", absCleanDst)
 					}
 					if absCleanDst == "" || absCleanDst == "/" {
-						return false, fmt.Errorf("Don't want to remove root!") // safety
+						return false, fmt.Errorf("don't want to remove root") // safety
 					}
 					log.Printf("syncCheckApply: Removing (force): %s", absCleanDst)
 					if err := os.Remove(absCleanDst); err != nil {
@@ -536,7 +596,7 @@ func (obj *FileRes) syncCheckApply(apply bool, src, dst string) (bool, error) {
 		absDst := fileInfo.AbsPath // absolute path (should get removed)
 		absCleanDst := path.Clean(absDst)
 		if absCleanDst == "" || absCleanDst == "/" {
-			return false, fmt.Errorf("Don't want to remove root!") // safety
+			return false, fmt.Errorf("don't want to remove root") // safety
 		}
 
 		// FIXME: respect obj.Recurse here...
@@ -594,7 +654,7 @@ func (obj *FileRes) contentCheckApply(apply bool) (checkOK bool, _ error) {
 
 		// apply portion
 		if obj.path == "" || obj.path == "/" {
-			return false, fmt.Errorf("Don't want to remove root!") // safety
+			return false, fmt.Errorf("don't want to remove root") // safety
 		}
 		log.Printf("contentCheckApply: Removing: %s", obj.path)
 		// FIXME: respect obj.Recurse here...
@@ -603,16 +663,16 @@ func (obj *FileRes) contentCheckApply(apply bool) (checkOK bool, _ error) {
 		return false, err             // either nil or not
 	}
 
+	if obj.isDir && obj.Source == "" {
+		return obj.dirCheckApply(apply)
+	}
+
 	// content is not defined, leave it alone...
 	if obj.Content == nil {
 		return true, nil
 	}
 
 	if obj.Source == "" { // do the obj.Content checks first...
-		if obj.isDir { // TODO: should we create an empty dir this way?
-			log.Fatal("XXX: Not implemented!") // XXX
-		}
-
 		bufferSrc := bytes.NewReader([]byte(*obj.Content))
 		sha256sum, checkOK, err := obj.fileCheckApply(apply, bufferSrc, obj.path, obj.sha256sum)
 		if sha256sum != "" { // empty values mean errored or didn't hash
@@ -635,6 +695,113 @@ func (obj *FileRes) contentCheckApply(apply bool) (checkOK bool, _ error) {
 	return checkOK, nil
 }
 
+// chmodCheckApply performs a CheckApply for the file permissions.
+func (obj *FileRes) chmodCheckApply(apply bool) (checkOK bool, _ error) {
+	log.Printf("%s[%s]: chmodCheckApply(%t)", obj.Kind(), obj.GetName(), apply)
+
+	if obj.State == "absent" {
+		// File is absent
+		return true, nil
+	}
+
+	if obj.Mode == "" {
+		// No mode specified, everything is ok
+		return true, nil
+	}
+
+	mode, err := obj.mode()
+
+	// If the file does not exist and we are in
+	// noop mode, do not throw an error.
+	if os.IsNotExist(err) && !apply {
+		return false, nil
+	}
+
+	if err != nil {
+		return false, err
+	}
+
+	st, err := os.Stat(obj.path)
+	if err != nil {
+		return false, err
+	}
+
+	// Nothing to do
+	if st.Mode() == mode {
+		return true, nil
+	}
+
+	// Not clean but don't apply
+	if !apply {
+		return false, nil
+	}
+
+	err = os.Chmod(obj.path, mode)
+	return false, err
+}
+
+// chownCheckApply performs a CheckApply for the file ownership.
+func (obj *FileRes) chownCheckApply(apply bool) (checkOK bool, _ error) {
+	var expectedUID, expectedGID int
+	log.Printf("%s[%s]: chownCheckApply(%t)", obj.Kind(), obj.GetName(), apply)
+
+	if obj.State == "absent" {
+		// File is absent or no owner specified
+		return true, nil
+	}
+
+	st, err := os.Stat(obj.path)
+
+	// If the file does not exist and we are in
+	// noop mode, do not throw an error.
+	if os.IsNotExist(err) && !apply {
+		return false, nil
+	}
+
+	if err != nil {
+		return false, err
+	}
+
+	stUnix, ok := st.Sys().(*syscall.Stat_t)
+	if !ok {
+		// Not unix
+		panic("No support for your platform")
+	}
+
+	if obj.Owner != "" {
+		expectedUID, err = obj.uid()
+		if err != nil {
+			return false, err
+		}
+	} else {
+		// Nothing specified, no changes to be made, expect same as actual
+		expectedUID = int(stUnix.Uid)
+	}
+
+	if obj.Group != "" {
+		expectedGID, err = obj.gid()
+		if err != nil {
+			return false, err
+		}
+	} else {
+		// Nothing specified, no changes to be made, expect same as actual
+		expectedGID = int(stUnix.Gid)
+	}
+
+	// Nothing to do
+	if int(stUnix.Uid) == expectedUID && int(stUnix.Gid) == expectedGID {
+		return true, nil
+	}
+
+	// Not clean, but don't apply
+	if !apply {
+		return false, nil
+	}
+
+	err = os.Chown(obj.path, expectedUID, expectedGID)
+	return false, err
+}
+
 // CheckApply checks the resource state and applies the resource if the bool
 // input is true. It returns error info and if the state check passed or not.
 func (obj *FileRes) CheckApply(apply bool) (checkOK bool, _ error) {
@@ -659,19 +826,17 @@ func (obj *FileRes) CheckApply(apply bool) (checkOK bool, _ error) {
 		checkOK = false
 	}
 
-	// TODO
-	//if c, err := obj.chmodCheckApply(apply); err != nil {
-	//	return false, err
-	//} else if !c {
-	//	checkOK = false
-	//}
+	if c, err := obj.chmodCheckApply(apply); err != nil {
+		return false, err
+	} else if !c {
+		checkOK = false
+	}
 
-	// TODO
-	//if c, err := obj.chownCheckApply(apply); err != nil {
-	//	return false, err
-	//} else if !c {
-	//	checkOK = false
-	//}
+	if c, err := obj.chownCheckApply(apply); err != nil {
+		return false, err
+	} else if !c {
+		checkOK = false
+	}
 
 	return checkOK, nil // w00t
 }
@@ -754,9 +919,9 @@ func (obj *FileRes) AutoEdges() AutoEdge {
 	}
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *FileRes) GetUIDs() []ResUID {
+func (obj *FileRes) UIDs() []ResUID {
 	x := &FileUID{
 		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
 		path:    obj.path,
@@ -787,7 +952,7 @@ func (obj *FileRes) Compare(res Res) bool {
 		if obj.Name != res.Name {
 			return false
 		}
-		if obj.path != res.Path {
+		if obj.path != res.path {
 			return false
 		}
 		if (obj.Content == nil) != (res.Content == nil) { // xor
@@ -821,3 +986,23 @@ func (obj *FileRes) CollectPattern(pattern string) {
 	// XXX: currently the pattern for files can only override the Dirname variable :P
 	obj.Dirname = pattern // XXX: simplistic for now
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *FileRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes FileRes // indirection to avoid infinite recursion
+
+	def := obj.Default()      // get the default
+	res, ok := def.(*FileRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to FileRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = FileRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/file_attrs.go

@@ -0,0 +1,43 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+// +build go1.7
+
+package resources
+
+import (
+	"os/user"
+	"strconv"
+
+	errwrap "github.com/pkg/errors"
+)
+
+// gid returns the group id for the group specified in the yaml file graph.
+// Caller should first check obj.Group is not empty
+func (obj *FileRes) gid() (int, error) {
+	g2, err2 := user.LookupGroupId(obj.Group)
+	if err2 == nil {
+		return strconv.Atoi(g2.Gid)
+	}
+
+	g, err := user.LookupGroup(obj.Group)
+	if err == nil {
+		return strconv.Atoi(g.Gid)
+	}
+
+	return -1, errwrap.Wrapf(err, "Group lookup error (%s)", obj.Group)
+}

resources/file_attrs_go1.6.go

@@ -0,0 +1,43 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+// +build !go1.7
+
+package resources
+
+import (
+	"strconv"
+
+	group "github.com/hnakamur/group"
+	errwrap "github.com/pkg/errors"
+)
+
+// gid returns the group id for the group specified in the yaml file graph.
+// Caller should first check obj.Group is not empty
+func (obj *FileRes) gid() (int, error) {
+	g2, err2 := group.LookupId(obj.Group)
+	if err2 == nil {
+		return strconv.Atoi(g2.Gid)
+	}
+
+	g, err := group.Lookup(obj.Group)
+	if err == nil {
+		return strconv.Atoi(g.Gid)
+	}
+
+	return -1, errwrap.Wrapf(err, "Group lookup error (%s)", obj.Group)
+}

resources/hostname.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -22,9 +22,7 @@ import (
 	"errors"
 	"fmt"
 	"log"
-	"time"
 
-	"github.com/purpleidea/mgmt/event"
 	"github.com/purpleidea/mgmt/util"
 
 	"github.com/godbus/dbus"
@@ -70,22 +68,26 @@ type HostnameRes struct {
 	conn *dbus.Conn
 }
 
-// NewHostnameRes is a constructor for this resource. It also calls Init() for you.
-func NewHostnameRes(name, staticHostname, transientHostname, prettyHostname string) (*HostnameRes, error) {
-	obj := &HostnameRes{
+// Default returns some sensible defaults for this resource.
+func (obj *HostnameRes) Default() Res {
+	return &HostnameRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		PrettyHostname:    prettyHostname,
-		StaticHostname:    staticHostname,
-		TransientHostname: transientHostname,
 	}
-	return obj, obj.Init()
+}
+
+// Validate if the params passed in are valid data.
+func (obj *HostnameRes) Validate() error {
+	if obj.PrettyHostname == "" && obj.StaticHostname == "" && obj.TransientHostname == "" {
+		return ErrResourceInsufficientParameters
+	}
+	return obj.BaseRes.Validate()
 }
 
 // Init runs some startup code for this resource.
 func (obj *HostnameRes) Init() error {
-	obj.BaseRes.kind = "Hostname"
+	obj.BaseRes.kind = "hostname"
 	if obj.PrettyHostname == "" {
 		obj.PrettyHostname = obj.Hostname
 	}
@@ -98,34 +100,8 @@ func (obj *HostnameRes) Init() error {
 	return obj.BaseRes.Init() // call base init, b/c we're overriding
 }
 
-// Validate if the params passed in are valid data.
-// FIXME: where should this get called ?
-func (obj *HostnameRes) Validate() error {
-	if obj.PrettyHostname == "" && obj.StaticHostname == "" && obj.TransientHostname == "" {
-		return ErrResourceInsufficientParameters
-	}
-	return nil
-}
-
 // Watch is the primary listener for this resource and it outputs events.
-func (obj *HostnameRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil // TODO: should this be an error?
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
-	}
-
+func (obj *HostnameRes) Watch() error {
 	// if we share the bus with others, we will get each others messages!!
 	bus, err := util.SystemBusPrivateUsable() // don't share the bus connection!
 	if err != nil {
@@ -142,42 +118,32 @@ func (obj *HostnameRes) Watch(processChan chan event.Event) error {
 	signals := make(chan *dbus.Signal, 10) // closed by dbus package
 	bus.Signal(signals)
 
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
 	var send = false // send event?
 
 	for {
-		obj.SetState(ResStateWatching) // reset
 		select {
 		case <-signals:
-			cuid.SetConverged(false)
 			send = true
 			obj.StateOK(false) // dirty
 
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
 			// we avoid sending events on unpause
-			if exit, _ := obj.ReadEvent(&event); exit {
-				return nil // exit
+			if exit, _ := obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
 			send = true
 			obj.StateOK(false) // dirty
-
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true) // converged!
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
-			startup = true // startup finished
 			send = false
-
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -266,9 +232,9 @@ func (obj *HostnameRes) AutoEdges() AutoEdge {
 	return nil
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *HostnameRes) GetUIDs() []ResUID {
+func (obj *HostnameRes) UIDs() []ResUID {
 	x := &HostnameUID{
 		BaseUID:           BaseUID{name: obj.GetName(), kind: obj.Kind()},
 		name:              obj.Name,
@@ -309,3 +275,23 @@ func (obj *HostnameRes) Compare(res Res) bool {
 	}
 	return true
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *HostnameRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes HostnameRes // indirection to avoid infinite recursion
+
+	def := obj.Default()          // get the default
+	res, ok := def.(*HostnameRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to HostnameRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = HostnameRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/kv.go

@@ -0,0 +1,304 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package resources
+
+import (
+	"encoding/gob"
+	"fmt"
+	"log"
+	"strconv"
+
+	errwrap "github.com/pkg/errors"
+)
+
+func init() {
+	gob.Register(&KVRes{})
+}
+
+// KVResSkipCmpStyle represents the different styles of comparison when using SkipLessThan.
+type KVResSkipCmpStyle int
+
+// These are the different allowed comparison styles. Most folks will want SkipCmpStyleInt.
+const (
+	SkipCmpStyleInt KVResSkipCmpStyle = iota
+	SkipCmpStyleString
+)
+
+// KVRes is a resource which writes a key/value pair into cluster wide storage.
+// It will ensure that the key is set to the requested value. The one exception
+// is that if you use the SkipLessThan parameter, then it will only replace the
+// stored value with the requested value if it is greater than that stored one.
+// This allows the KV resource to be used in fast acting, finite state machines
+// which have monotonically increasing state values that represent progression.
+// The one exception is that when this resource receives a refresh signal, then
+// it will set the value to be the exact one if they are not identical already.
+type KVRes struct {
+	BaseRes      `yaml:",inline"`
+	Key          string            `yaml:"key"`          // key to set
+	Value        *string           `yaml:"value"`        // value to set (nil to delete)
+	SkipLessThan bool              `yaml:"skiplessthan"` // skip updates as long as stored value is greater
+	SkipCmpStyle KVResSkipCmpStyle `yaml:"skipcmpstyle"` // how to do the less than cmp
+	// TODO: does it make sense to have different backends here? (eg: local)
+}
+
+// Default returns some sensible defaults for this resource.
+func (obj *KVRes) Default() Res {
+	return &KVRes{
+		BaseRes: BaseRes{
+			MetaParams: DefaultMetaParams, // force a default
+		},
+	}
+}
+
+// Validate if the params passed in are valid data.
+// FIXME: This will catch most issues unless data is passed in after Init with
+// the Send/Recv mechanism. Should the engine re-call Validate after Send/Recv?
+func (obj *KVRes) Validate() error {
+	if obj.Key == "" {
+		return fmt.Errorf("key must not be empty")
+	}
+	if obj.SkipLessThan {
+		if obj.SkipCmpStyle != SkipCmpStyleInt && obj.SkipCmpStyle != SkipCmpStyleString {
+			return fmt.Errorf("the SkipCmpStyle of %v is invalid", obj.SkipCmpStyle)
+		}
+
+		if v := obj.Value; obj.SkipCmpStyle == SkipCmpStyleInt && v != nil {
+			if _, err := strconv.Atoi(*v); err != nil {
+				return fmt.Errorf("the set value of %v can't convert to int", v)
+			}
+		}
+	}
+	return obj.BaseRes.Validate()
+}
+
+// Init initializes the resource.
+func (obj *KVRes) Init() error {
+	obj.BaseRes.kind = "kv"
+	return obj.BaseRes.Init() // call base init, b/c we're overriding
+}
+
+// Watch is the primary listener for this resource and it outputs events.
+func (obj *KVRes) Watch() error {
+
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
+	ch := obj.Data().World.StrWatch(obj.Key) // get possible events!
+
+	var send = false // send event?
+	var exit *error
+	for {
+		select {
+		// NOTE: this part is very similar to the file resource code
+		case err, ok := <-ch:
+			if !ok { // channel shutdown
+				return nil
+			}
+			if err != nil {
+				return errwrap.Wrapf(err, "unknown %s[%s] watcher error", obj.Kind(), obj.GetName())
+			}
+			if obj.Data().Debug {
+				log.Printf("%s[%s]: Event!", obj.Kind(), obj.GetName())
+			}
+			send = true
+			obj.StateOK(false) // dirty
+
+		case event := <-obj.Events():
+			// we avoid sending events on unpause
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
+			}
+		}
+
+		// do all our event sending all together to avoid duplicate msgs
+		if send {
+			send = false
+			obj.Event()
+		}
+	}
+}
+
+// lessThanCheck checks for less than validity.
+func (obj *KVRes) lessThanCheck(value string) (checkOK bool, err error) {
+
+	v := *obj.Value
+	if value == v { // redundant check for safety
+		return true, nil
+	}
+
+	var refresh = obj.Refresh()       // do we have a pending reload to apply?
+	if !obj.SkipLessThan || refresh { // update lessthan on refresh
+		return false, nil
+	}
+
+	switch obj.SkipCmpStyle {
+	case SkipCmpStyleInt:
+		intValue, err := strconv.Atoi(value)
+		if err != nil {
+			// NOTE: We don't error here since we're going to write
+			// over the value anyways. It could be from an old run!
+			return false, nil // value is bad (old/corrupt), fix it
+		}
+		if vint, err := strconv.Atoi(v); err != nil {
+			return false, errwrap.Wrapf(err, "can't convert %v to int", v)
+		} else if vint < intValue {
+			return true, nil
+		}
+
+	case SkipCmpStyleString:
+		if v < value { // weird way to cmp, but valid
+			return true, nil
+		}
+
+	default:
+		return false, fmt.Errorf("unmatches SkipCmpStyle style %v", obj.SkipCmpStyle)
+	}
+
+	return false, nil
+}
+
+// CheckApply method for Password resource. Does nothing, returns happy!
+func (obj *KVRes) CheckApply(apply bool) (checkOK bool, err error) {
+	log.Printf("%s[%s]: CheckApply(%t)", obj.Kind(), obj.GetName(), apply)
+
+	if val, exists := obj.Recv["Value"]; exists && val.Changed {
+		// if we received on Value, and it changed, wooo, nothing to do.
+		log.Printf("CheckApply: `Value` was updated!")
+	}
+
+	hostname := obj.Data().Hostname // me
+	keyMap, err := obj.Data().World.StrGet(obj.Key)
+	if err != nil {
+		return false, errwrap.Wrapf(err, "check error during StrGet")
+	}
+
+	if value, ok := keyMap[hostname]; ok && obj.Value != nil {
+		if value == *obj.Value {
+			return true, nil
+		}
+
+		if c, err := obj.lessThanCheck(value); err != nil {
+			return false, err
+		} else if c {
+			return true, nil
+		}
+
+	} else if !ok && obj.Value == nil {
+		return true, nil // nothing to delete, we're good!
+
+	} else if ok && obj.Value == nil { // delete
+		err := obj.Data().World.StrDel(obj.Key)
+		return false, errwrap.Wrapf(err, "apply error during StrDel")
+	}
+
+	if !apply {
+		return false, nil
+	}
+
+	if err := obj.Data().World.StrSet(obj.Key, *obj.Value); err != nil {
+		return false, errwrap.Wrapf(err, "apply error during StrSet")
+	}
+
+	return false, nil
+}
+
+// KVUID is the UID struct for KVRes.
+type KVUID struct {
+	BaseUID
+	name string
+}
+
+// AutoEdges returns the AutoEdge interface. In this case no autoedges are used.
+func (obj *KVRes) AutoEdges() AutoEdge {
+	return nil
+}
+
+// UIDs includes all params to make a unique identification of this object.
+// Most resources only return one, although some resources can return multiple.
+func (obj *KVRes) UIDs() []ResUID {
+	x := &KVUID{
+		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
+		name:    obj.Name,
+	}
+	return []ResUID{x}
+}
+
+// GroupCmp returns whether two resources can be grouped together or not.
+func (obj *KVRes) GroupCmp(r Res) bool {
+	_, ok := r.(*KVRes)
+	if !ok {
+		return false
+	}
+	return false // TODO: this is doable!
+	// TODO: it could be useful to group our writes and watches!
+}
+
+// Compare two resources and return if they are equivalent.
+func (obj *KVRes) Compare(res Res) bool {
+	switch res.(type) {
+	// we can only compare KVRes to others of the same resource
+	case *KVRes:
+		res := res.(*KVRes)
+		if !obj.BaseRes.Compare(res) { // call base Compare
+			return false
+		}
+
+		if obj.Key != res.Key {
+			return false
+		}
+		if (obj.Value == nil) != (res.Value == nil) { // xor
+			return false
+		}
+		if obj.Value != nil && res.Value != nil {
+			if *obj.Value != *res.Value { // compare the strings
+				return false
+			}
+		}
+		if obj.SkipLessThan != res.SkipLessThan {
+			return false
+		}
+		if obj.SkipCmpStyle != res.SkipCmpStyle {
+			return false
+		}
+	default:
+		return false
+	}
+	return true
+}
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *KVRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes KVRes // indirection to avoid infinite recursion
+
+	def := obj.Default()    // get the default
+	res, ok := def.(*KVRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to KVRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = KVRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/msg.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -23,9 +23,6 @@ import (
 	"log"
 	"regexp"
 	"strings"
-	"time"
-
-	"github.com/purpleidea/mgmt/event"
 
 	"github.com/coreos/go-systemd/journal"
 )
@@ -53,45 +50,33 @@ type MsgUID struct {
 	body string
 }
 
-// NewMsgRes is a constructor for this resource.
-func NewMsgRes(name, body, priority string, journal, syslog bool, fields map[string]string) (*MsgRes, error) {
-	message := name
-	if body != "" {
-		message = body
-	}
-
-	obj := &MsgRes{
+// Default returns some sensible defaults for this resource.
+func (obj *MsgRes) Default() Res {
+	return &MsgRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		Body:     message,
-		Priority: priority,
-		Fields:   fields,
-		Journal:  journal,
-		Syslog:   syslog,
 	}
-
-	return obj, obj.Init()
 }
 
-// Init runs some startup code for this resource.
-func (obj *MsgRes) Init() error {
-	obj.BaseRes.kind = "Msg"
-	return obj.BaseRes.Init() // call base init, b/c we're overrriding
-}
-
-// Validate the params that are passed to MsgRes
+// Validate the params that are passed to MsgRes.
 func (obj *MsgRes) Validate() error {
 	invalidCharacters := regexp.MustCompile("[^a-zA-Z0-9_]")
 	for field := range obj.Fields {
 		if invalidCharacters.FindString(field) != "" {
-			return fmt.Errorf("Invalid character in field %s.", field)
+			return fmt.Errorf("invalid character in field %s", field)
 		}
 		if strings.HasPrefix(field, "_") {
-			return fmt.Errorf("Fields cannot begin with _.")
+			return fmt.Errorf("fields cannot begin with _")
 		}
 	}
-	return nil
+	return obj.BaseRes.Validate()
+}
+
+// Init runs some startup code for this resource.
+func (obj *MsgRes) Init() error {
+	obj.BaseRes.kind = "msg"
+	return obj.BaseRes.Init() // call base init, b/c we're overrriding
 }
 
 // isAllStateOK derives a compound state from all internal cache flags that apply to this resource.
@@ -135,54 +120,27 @@ func (obj *MsgRes) journalPriority() journal.Priority {
 }
 
 // Watch is the primary listener for this resource and it outputs events.
-func (obj *MsgRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
+func (obj *MsgRes) Watch() error {
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
 	}
 
 	var send = false // send event?
-	var exit = false
+	var exit *error
 	for {
-		obj.SetState(ResStateWatching) // reset
 		select {
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
 			// we avoid sending events on unpause
-			if exit, send = obj.ReadEvent(&event); exit {
-				return nil // exit
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
-
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true) // converged!
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
-			startup = true // startup finished
 			send = false
-			// only do this on certain types of events
-			//obj.isStateOK = false // something made state dirty
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -232,9 +190,9 @@ func (obj *MsgRes) CheckApply(apply bool) (bool, error) {
 	return false, nil
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *MsgRes) GetUIDs() []ResUID {
+func (obj *MsgRes) UIDs() []ResUID {
 	x := &MsgUID{
 		BaseUID: BaseUID{
 			name: obj.GetName(),
@@ -277,3 +235,23 @@ func (obj *MsgRes) Compare(res Res) bool {
 	}
 	return true
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *MsgRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes MsgRes // indirection to avoid infinite recursion
+
+	def := obj.Default()     // get the default
+	res, ok := def.(*MsgRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to MsgRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = MsgRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/noop.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -19,10 +19,8 @@ package resources
 
 import (
 	"encoding/gob"
+	"fmt"
 	"log"
-	"time"
-
-	"github.com/purpleidea/mgmt/event"
 )
 
 func init() {
@@ -35,76 +33,48 @@ type NoopRes struct {
 	Comment string `yaml:"comment"` // extra field for example purposes
 }
 
-// NewNoopRes is a constructor for this resource. It also calls Init() for you.
-func NewNoopRes(name string) (*NoopRes, error) {
-	obj := &NoopRes{
+// Default returns some sensible defaults for this resource.
+func (obj *NoopRes) Default() Res {
+	return &NoopRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		Comment: "",
 	}
-	return obj, obj.Init()
+}
+
+// Validate if the params passed in are valid data.
+func (obj *NoopRes) Validate() error {
+	return obj.BaseRes.Validate()
 }
 
 // Init runs some startup code for this resource.
 func (obj *NoopRes) Init() error {
-	obj.BaseRes.kind = "Noop"
+	obj.BaseRes.kind = "noop"
 	return obj.BaseRes.Init() // call base init, b/c we're overriding
 }
 
-// Validate if the params passed in are valid data.
-// FIXME: where should this get called ?
-func (obj *NoopRes) Validate() error {
-	return nil
-}
-
 // Watch is the primary listener for this resource and it outputs events.
-func (obj *NoopRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil // TODO: should this be an error?
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
+func (obj *NoopRes) Watch() error {
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
 	}
 
 	var send = false // send event?
-	var exit = false
+	var exit *error
 	for {
-		obj.SetState(ResStateWatching) // reset
 		select {
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
 			// we avoid sending events on unpause
-			if exit, send = obj.ReadEvent(&event); exit {
-				return nil // exit
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
-
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true) // converged!
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
-			startup = true // startup finished
 			send = false
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -128,9 +98,9 @@ func (obj *NoopRes) AutoEdges() AutoEdge {
 	return nil
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *NoopRes) GetUIDs() []ResUID {
+func (obj *NoopRes) UIDs() []ResUID {
 	x := &NoopUID{
 		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
 		name:    obj.Name,
@@ -169,3 +139,23 @@ func (obj *NoopRes) Compare(res Res) bool {
 	}
 	return true
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *NoopRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes NoopRes // indirection to avoid infinite recursion
+
+	def := obj.Default()      // get the default
+	res, ok := def.(*NoopRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to NoopRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = NoopRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/nspawn.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -22,9 +22,7 @@ import (
 	"errors"
 	"fmt"
 	"log"
-	"time"
 
-	"github.com/purpleidea/mgmt/event"
 	"github.com/purpleidea/mgmt/util"
 
 	systemdUtil "github.com/coreos/go-systemd/util"
@@ -58,7 +56,34 @@ type NspawnRes struct {
 	svc *SvcRes
 }
 
-// Init runs some startup code for this resource
+// Default returns some sensible defaults for this resource.
+func (obj *NspawnRes) Default() Res {
+	return &NspawnRes{
+		BaseRes: BaseRes{
+			MetaParams: DefaultMetaParams, // force a default
+		},
+		State: running,
+	}
+}
+
+// Validate if the params passed in are valid data.
+func (obj *NspawnRes) Validate() error {
+	// TODO: validStates should be an enum!
+	validStates := map[string]struct{}{
+		stopped: {},
+		running: {},
+	}
+	if _, exists := validStates[obj.State]; !exists {
+		return fmt.Errorf("Invalid State: %s", obj.State)
+	}
+
+	if err := obj.svc.Validate(); err != nil { // composite resource
+		return errwrap.Wrapf(err, "validate failed for embedded svc")
+	}
+	return obj.BaseRes.Validate()
+}
+
+// Init runs some startup code for this resource.
 func (obj *NspawnRes) Init() error {
 	var serviceName = fmt.Sprintf(nspawnServiceTmpl, obj.GetName())
 	obj.svc = &SvcRes{}
@@ -67,61 +92,21 @@ func (obj *NspawnRes) Init() error {
 	if err := obj.svc.Init(); err != nil {
 		return err
 	}
-	obj.BaseRes.kind = "Nspawn"
+	obj.BaseRes.kind = "nspawn"
 	return obj.BaseRes.Init()
 }
 
-// NewNspawnRes is the constructor for this resource
-func NewNspawnRes(name string, state string) (*NspawnRes, error) {
-	obj := &NspawnRes{
-		BaseRes: BaseRes{
-			Name: name,
-		},
-		State: state,
-	}
-	return obj, obj.Init()
-}
-
-// Validate params
-func (obj *NspawnRes) Validate() error {
-	validStates := map[string]struct{}{
-		stopped: {},
-		running: {},
-	}
-	if _, exists := validStates[obj.State]; !exists {
-		return fmt.Errorf("Invalid State: %s", obj.State)
-	}
-	return obj.svc.Validate()
-}
-
 // Watch for state changes and sends a message to the bus if there is a change
-func (obj *NspawnRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-		}
-		// 1/2 the resolution of converged timeout
-		return time.After(time.Duration(500) * time.Millisecond)
-	}
-
+func (obj *NspawnRes) Watch() error {
 	// this resource depends on systemd ensure that it's running
 	if !systemdUtil.IsRunningSystemd() {
-		return fmt.Errorf("Systemd is not running.")
+		return fmt.Errorf("systemd is not running")
 	}
 
 	// create a private message bus
 	bus, err := util.SystemBusPrivateUsable()
 	if err != nil {
-		return errwrap.Wrapf(err, "Failed to connect to bus")
+		return errwrap.Wrapf(err, "failed to connect to bus")
 	}
 
 	// add a match rule to match messages going through the message bus
@@ -135,11 +120,15 @@ func (obj *NspawnRes) Watch(processChan chan event.Event) error {
 	buschan := make(chan *dbus.Signal, 10)
 	bus.Signal(buschan)
 
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
 	var send = false
-	var exit = false
+	var exit *error
 
 	for {
-		obj.SetState(ResStateWatching)
 		select {
 		case event := <-buschan:
 			// process org.freedesktop.machine1 events for this resource's name
@@ -150,35 +139,22 @@ func (obj *NspawnRes) Watch(processChan chan event.Event) error {
 				} else if event.Name == machineRemoved {
 					log.Printf("%s[%s]: Machine stopped", obj.Kind(), obj.GetName())
 				} else {
-					return fmt.Errorf("Unknown event: %s", event.Name)
+					return fmt.Errorf("unknown event: %s", event.Name)
 				}
 				send = true
 				obj.StateOK(false) // dirty
 			}
 
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
-			if exit, send = obj.ReadEvent(&event); exit {
-				return nil // exit
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
-
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true) // converged!
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
-			obj.StateOK(false) // dirty
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
-			startup = true // startup finished
 			send = false
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -189,13 +165,13 @@ func (obj *NspawnRes) Watch(processChan chan event.Event) error {
 func (obj *NspawnRes) CheckApply(apply bool) (checkOK bool, err error) {
 	// this resource depends on systemd ensure that it's running
 	if !systemdUtil.IsRunningSystemd() {
-		return false, errors.New("Systemd is not running.")
+		return false, errors.New("systemd is not running")
 	}
 
 	// connect to org.freedesktop.machine1.Manager
 	conn, err := machined.New()
 	if err != nil {
-		return false, errwrap.Wrapf(err, "Failed to connect to dbus")
+		return false, errwrap.Wrapf(err, "failed to connect to dbus")
 	}
 
 	// compare the current state with the desired state and perform the
@@ -213,7 +189,7 @@ func (obj *NspawnRes) CheckApply(apply bool) (checkOK bool, err error) {
 		// error if we need the image ignore if we don't
 		if _, err = conn.GetImage(obj.GetName()); err != nil && obj.State != stopped {
 			return false, fmt.Errorf(
-				"No machine nor image named '%s'",
+				"no machine nor image named '%s'",
 				obj.GetName())
 		}
 	}
@@ -243,7 +219,7 @@ func (obj *NspawnRes) CheckApply(apply bool) (checkOK bool, err error) {
 		log.Printf("%s[%s]: Starting machine", obj.Kind(), obj.GetName())
 		// assume state had to be changed at this point, ignore checkOK
 		if _, err := obj.svc.CheckApply(apply); err != nil {
-			return false, errwrap.Wrapf(err, "Nested svc failed")
+			return false, errwrap.Wrapf(err, "nested svc failed")
 		}
 	}
 	if obj.State == stopped {
@@ -251,7 +227,7 @@ func (obj *NspawnRes) CheckApply(apply bool) (checkOK bool, err error) {
 		// org.freedesktop.machine1.Manager.KillMachine
 		log.Printf("%s[%s]: Stopping machine", obj.Kind(), obj.GetName())
 		if err := conn.TerminateMachine(obj.GetName()); err != nil {
-			return false, errwrap.Wrapf(err, "Failed to stop machine")
+			return false, errwrap.Wrapf(err, "failed to stop machine")
 		}
 	}
 
@@ -277,14 +253,14 @@ func (obj *NspawnUID) IFF(uid ResUID) bool {
 	return obj.name == res.name
 }
 
-// GetUIDs includes all params to make a unique identification of this object
+// UIDs includes all params to make a unique identification of this object
 // most resources only return one although some resources can return multiple
-func (obj *NspawnRes) GetUIDs() []ResUID {
+func (obj *NspawnRes) UIDs() []ResUID {
 	x := &NspawnUID{
 		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
 		name:    obj.Name, // svc name
 	}
-	return append([]ResUID{x}, obj.svc.GetUIDs()...)
+	return append([]ResUID{x}, obj.svc.UIDs()...)
 }
 
 // GroupCmp returns whether two resources can be grouped together or not
@@ -321,3 +297,23 @@ func (obj *NspawnRes) Compare(res Res) bool {
 func (obj *NspawnRes) AutoEdges() AutoEdge {
 	return nil
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *NspawnRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes NspawnRes // indirection to avoid infinite recursion
+
+	def := obj.Default()        // get the default
+	res, ok := def.(*NspawnRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to NspawnRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = NspawnRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/packagekit/packagekit.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -833,7 +833,7 @@ func FilterPackageIDs(m map[string]*PkPackageIDActionData, packages []string) ([
 		obj, ok := m[k] // lookup single package
 		// package doesn't exist, this is an error!
 		if !ok || !obj.Found || obj.PackageID == "" {
-			return nil, fmt.Errorf("Can't find package named '%s'.", k)
+			return nil, fmt.Errorf("can't find package named '%s'", k)
 		}
 		result = append(result, obj.PackageID)
 	}
@@ -848,7 +848,7 @@ func FilterState(m map[string]*PkPackageIDActionData, packages []string, state s
 		obj, ok := m[k] // lookup single package
 		// package doesn't exist, this is an error!
 		if !ok || !obj.Found {
-			return nil, fmt.Errorf("Can't find package named '%s'.", k)
+			return nil, fmt.Errorf("can't find package named '%s'", k)
 		}
 		var b bool
 		if state == "installed" {
@@ -865,7 +865,7 @@ func FilterState(m map[string]*PkPackageIDActionData, packages []string, state s
 		result[k] = b // save
 	}
 	if len(pkgs) > 0 {
-		err = fmt.Errorf("Can't filter non-boolean state on: %v!", strings.Join(pkgs, ","))
+		err = fmt.Errorf("can't filter non-boolean state on: %v", strings.Join(pkgs, ","))
 	}
 	return result, err
 }
@@ -877,7 +877,7 @@ func FilterPackageState(m map[string]*PkPackageIDActionData, packages []string,
 		obj, ok := m[k] // lookup single package
 		// package doesn't exist, this is an error!
 		if !ok || !obj.Found {
-			return nil, fmt.Errorf("Can't find package named '%s'.", k)
+			return nil, fmt.Errorf("can't find package named '%s'", k)
 		}
 		b := false
 		if state == "installed" && obj.Installed {

resources/password.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -27,9 +27,7 @@ import (
 	"os"
 	"path"
 	"strings"
-	"time"
 
-	"github.com/purpleidea/mgmt/event"
 	"github.com/purpleidea/mgmt/recwatch"
 
 	errwrap "github.com/pkg/errors"
@@ -57,21 +55,25 @@ type PasswordRes struct {
 	recWatcher *recwatch.RecWatcher
 }
 
-// NewPasswordRes is a constructor for this resource. It also calls Init() for you.
-func NewPasswordRes(name string, length uint16) (*PasswordRes, error) {
-	obj := &PasswordRes{
+// Default returns some sensible defaults for this resource.
+func (obj *PasswordRes) Default() Res {
+	return &PasswordRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		Length: length,
+		Length: 64, // safe default
 	}
-	return obj, obj.Init()
+}
+
+// Validate if the params passed in are valid data.
+func (obj *PasswordRes) Validate() error {
+	return obj.BaseRes.Validate()
 }
 
 // Init generates a new password for this resource if one was not provided. It
 // will save this into a local file. It will load it back in from previous runs.
 func (obj *PasswordRes) Init() error {
-	obj.BaseRes.kind = "Password" // must be set before using VarDir
+	obj.BaseRes.kind = "password" // must be set before using VarDir
 
 	dir, err := obj.VarDir("")
 	if err != nil {
@@ -82,12 +84,6 @@ func (obj *PasswordRes) Init() error {
 	return obj.BaseRes.Init() // call base init, b/c we're overriding
 }
 
-// Validate if the params passed in are valid data.
-// FIXME: where should this get called ?
-func (obj *PasswordRes) Validate() error {
-	return nil
-}
-
 func (obj *PasswordRes) read() (string, error) {
 	file, err := os.Open(obj.path) // open a handle to read the file
 	if err != nil {
@@ -148,11 +144,11 @@ func (obj *PasswordRes) check(value string) error {
 		return nil
 	}
 	if !obj.Saved && length != 0 { // should have no stored password
-		return fmt.Errorf("Expected empty token only!")
+		return fmt.Errorf("expected empty token only")
 	}
 
 	if length != obj.Length {
-		return fmt.Errorf("String length is not %d", obj.Length)
+		return fmt.Errorf("string length is not %d", obj.Length)
 	}
 Loop:
 	for i := uint16(0); i < length; i++ {
@@ -162,30 +158,13 @@ Loop:
 			}
 		}
 		// we couldn't find that character, so error!
-		return fmt.Errorf("Invalid character `%s`", string(value[i]))
+		return fmt.Errorf("invalid character `%s`", string(value[i]))
 	}
 	return nil
 }
 
 // Watch is the primary listener for this resource and it outputs events.
-func (obj *PasswordRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil // TODO: should this be an error?
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
-	}
-
+func (obj *PasswordRes) Watch() error {
 	var err error
 	obj.recWatcher, err = recwatch.NewRecWatcher(obj.path, false)
 	if err != nil {
@@ -193,46 +172,37 @@ func (obj *PasswordRes) Watch(processChan chan event.Event) error {
 	}
 	defer obj.recWatcher.Close()
 
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
 	var send = false // send event?
-	var exit = false
+	var exit *error
 	for {
-		obj.SetState(ResStateWatching) // reset
 		select {
 		// NOTE: this part is very similar to the file resource code
 		case event, ok := <-obj.recWatcher.Events():
 			if !ok { // channel shutdown
 				return nil
 			}
-			cuid.SetConverged(false)
 			if err := event.Error; err != nil {
-				return errwrap.Wrapf(err, "Unknown %s[%s] watcher error", obj.Kind(), obj.GetName())
+				return errwrap.Wrapf(err, "unknown %s[%s] watcher error", obj.Kind(), obj.GetName())
 			}
 			send = true
 			obj.StateOK(false) // dirty
 
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
 			// we avoid sending events on unpause
-			if exit, send = obj.ReadEvent(&event); exit {
-				return nil // exit
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
-
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true) // converged!
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
-			startup = true // startup finished
 			send = false
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -329,9 +299,9 @@ func (obj *PasswordRes) AutoEdges() AutoEdge {
 	return nil
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *PasswordRes) GetUIDs() []ResUID {
+func (obj *PasswordRes) UIDs() []ResUID {
 	x := &PasswordUID{
 		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
 		name:    obj.Name,
@@ -379,3 +349,23 @@ func (obj *PasswordRes) Compare(res Res) bool {
 	}
 	return true
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *PasswordRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes PasswordRes // indirection to avoid infinite recursion
+
+	def := obj.Default()          // get the default
+	res, ok := def.(*PasswordRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to PasswordRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = PasswordRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/pkg.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -23,9 +23,7 @@ import (
 	"log"
 	"path"
 	"strings"
-	"time"
 
-	"github.com/purpleidea/mgmt/event"
 	"github.com/purpleidea/mgmt/resources/packagekit"
 	"github.com/purpleidea/mgmt/util"
 
@@ -47,48 +45,53 @@ type PkgRes struct {
 	fileList []string // FIXME: update if pkg changes
 }
 
-// NewPkgRes is a constructor for this resource. It also calls Init() for you.
-func NewPkgRes(name, state string, allowuntrusted, allownonfree, allowunsupported bool) (*PkgRes, error) {
-	obj := &PkgRes{
+// Default returns some sensible defaults for this resource.
+func (obj *PkgRes) Default() Res {
+	return &PkgRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		State:            state,
-		AllowUntrusted:   allowuntrusted,
-		AllowNonFree:     allownonfree,
-		AllowUnsupported: allowunsupported,
+		State: "installed", // i think this is preferable to "latest"
 	}
-	return obj, obj.Init()
+}
+
+// Validate checks if the resource data structure was populated correctly.
+func (obj *PkgRes) Validate() error {
+	if obj.State == "" {
+		return fmt.Errorf("state cannot be empty")
+	}
+
+	return obj.BaseRes.Validate()
 }
 
 // Init runs some startup code for this resource.
 func (obj *PkgRes) Init() error {
-	obj.BaseRes.kind = "Pkg"
+	obj.BaseRes.kind = "pkg"
 	if err := obj.BaseRes.Init(); err != nil { // call base init, b/c we're overriding
 		return err
 	}
 
 	bus := packagekit.NewBus()
 	if bus == nil {
-		return fmt.Errorf("Can't connect to PackageKit bus.")
+		return fmt.Errorf("can't connect to PackageKit bus")
 	}
 	defer bus.Close()
 
 	result, err := obj.pkgMappingHelper(bus)
 	if err != nil {
-		return errwrap.Wrapf(err, "The pkgMappingHelper failed")
+		return errwrap.Wrapf(err, "the pkgMappingHelper failed")
 	}
 
 	data, ok := result[obj.Name] // lookup single package (init does just one)
 	// package doesn't exist, this is an error!
 	if !ok || !data.Found {
-		return fmt.Errorf("Can't find package named '%s'.", obj.Name)
+		return fmt.Errorf("can't find package named '%s'", obj.Name)
 	}
 
 	packageIDs := []string{data.PackageID} // just one for now
 	filesMap, err := bus.GetFilesByPackageID(packageIDs)
 	if err != nil {
-		return errwrap.Wrapf(err, "Can't run GetFilesByPackageID")
+		return errwrap.Wrapf(err, "can't run GetFilesByPackageID")
 	}
 	if files, ok := filesMap[data.PackageID]; ok {
 		obj.fileList = util.DirifyFileList(files, false)
@@ -96,61 +99,37 @@ func (obj *PkgRes) Init() error {
 	return nil
 }
 
-// Validate checks if the resource data structure was populated correctly.
-func (obj *PkgRes) Validate() error {
-	if obj.State == "" {
-		return fmt.Errorf("State cannot be empty!")
-	}
-
-	return nil
-}
-
 // Watch is the primary listener for this resource and it outputs events.
 // It uses the PackageKit UpdatesChanged signal to watch for changes.
 // TODO: https://github.com/hughsie/PackageKit/issues/109
 // TODO: https://github.com/hughsie/PackageKit/issues/110
-func (obj *PkgRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
-	}
-
+func (obj *PkgRes) Watch() error {
 	bus := packagekit.NewBus()
 	if bus == nil {
-		return fmt.Errorf("Can't connect to PackageKit bus.")
+		return fmt.Errorf("can't connect to PackageKit bus")
 	}
 	defer bus.Close()
 
 	ch, err := bus.WatchChanges()
 	if err != nil {
-		return errwrap.Wrapf(err, "Error adding signal match")
+		return errwrap.Wrapf(err, "error adding signal match")
+	}
+
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
 	}
 
 	var send = false // send event?
-	var exit = false
+	var exit *error
 
 	for {
 		if obj.debug {
 			log.Printf("%s: Watching...", obj.fmtNames(obj.getNames()))
 		}
 
-		obj.SetState(ResStateWatching) // reset
 		select {
 		case event := <-ch:
-			cuid.SetConverged(false)
-
 			// FIXME: ask packagekit for info on what packages changed
 			if obj.debug {
 				log.Printf("%s: Event: %v", obj.fmtNames(obj.getNames()), event.Name)
@@ -166,29 +145,17 @@ func (obj *PkgRes) Watch(processChan chan event.Event) error {
 			obj.StateOK(false) // dirty
 
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
-			if exit, send = obj.ReadEvent(&event); exit {
-				return nil // exit
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
 			//obj.StateOK(false) // these events don't invalidate state
 
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true) // converged!
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
-			obj.StateOK(false) // dirty
 		}
 
 		// do all our event sending all together to avoid duplicate msgs
 		if send {
-			startup = true // startup finished
 			send = false
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -222,7 +189,7 @@ func (obj *PkgRes) groupMappingHelper() map[string]string {
 		for _, x := range g {
 			pkg, ok := x.(*PkgRes) // convert from Res
 			if !ok {
-				log.Fatalf("Grouped member %v is not a %s", x, obj.Kind())
+				log.Fatalf("grouped member %v is not a %s", x, obj.Kind())
 			}
 			result[pkg.Name] = pkg.State
 		}
@@ -262,13 +229,13 @@ func (obj *PkgRes) CheckApply(apply bool) (checkOK bool, err error) {
 
 	bus := packagekit.NewBus()
 	if bus == nil {
-		return false, fmt.Errorf("Can't connect to PackageKit bus.")
+		return false, fmt.Errorf("can't connect to PackageKit bus")
 	}
 	defer bus.Close()
 
 	result, err := obj.pkgMappingHelper(bus)
 	if err != nil {
-		return false, errwrap.Wrapf(err, "The pkgMappingHelper failed")
+		return false, errwrap.Wrapf(err, "the pkgMappingHelper failed")
 	}
 
 	packageMap := obj.groupMappingHelper() // map[string]string
@@ -281,7 +248,7 @@ func (obj *PkgRes) CheckApply(apply bool) (checkOK bool, err error) {
 	// eventually we might be able to drop this constraint!
 	states, err := packagekit.FilterState(result, packageList, obj.State)
 	if err != nil {
-		return false, errwrap.Wrapf(err, "The FilterState method failed")
+		return false, errwrap.Wrapf(err, "the FilterState method failed")
 	}
 	data, _ := result[obj.Name] // if above didn't error, we won't either!
 	validState := util.BoolMapTrue(util.BoolMapValues(states))
@@ -373,7 +340,7 @@ type PkgResAutoEdges struct {
 // Next returns the next automatic edge.
 func (obj *PkgResAutoEdges) Next() []ResUID {
 	if obj.testIsNext {
-		log.Fatal("Expecting a call to Test()")
+		log.Fatal("expecting a call to Test()")
 	}
 	obj.testIsNext = true // set after all the errors paths are past
 
@@ -401,13 +368,13 @@ func (obj *PkgResAutoEdges) Next() []ResUID {
 // Test gets results of the earlier Next() call, & returns if we should continue!
 func (obj *PkgResAutoEdges) Test(input []bool) bool {
 	if !obj.testIsNext {
-		log.Fatal("Expecting a call to Next()")
+		log.Fatal("expecting a call to Next()")
 	}
 
 	// ack the svcUID's...
 	if x := obj.svcUIDs; len(x) > 0 {
 		if y := len(x); y != len(input) {
-			log.Fatalf("Expecting %d value(s)!", y)
+			log.Fatalf("expecting %d value(s)", y)
 		}
 		obj.svcUIDs = []ResUID{} // empty
 		obj.testIsNext = false
@@ -416,7 +383,7 @@ func (obj *PkgResAutoEdges) Test(input []bool) bool {
 
 	count := len(obj.fileList)
 	if count != len(input) {
-		log.Fatalf("Expecting %d value(s)!", count)
+		log.Fatalf("expecting %d value(s)", count)
 	}
 	obj.testIsNext = false // set after all the errors paths are past
 
@@ -479,9 +446,9 @@ func (obj *PkgRes) AutoEdges() AutoEdge {
 	}
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *PkgRes) GetUIDs() []ResUID {
+func (obj *PkgRes) UIDs() []ResUID {
 	x := &PkgUID{
 		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
 		name:    obj.Name,
@@ -561,3 +528,23 @@ func ReturnSvcInFileList(fileList []string) []string {
 	}
 	return result
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *PkgRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes PkgRes // indirection to avoid infinite recursion
+
+	def := obj.Default()     // get the default
+	res, ok := def.(*PkgRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to PkgRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = PkgRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/refresh.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify

resources/resources.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -19,19 +19,23 @@
 package resources
 
 import (
-	"bytes"
-	"encoding/base64"
-	"encoding/gob"
 	"fmt"
 	"log"
+	"math"
 	"os"
 	"path"
+	"sort"
+	"sync"
+	"time"
 
 	// TODO: should each resource be a sub-package?
 	"github.com/purpleidea/mgmt/converger"
 	"github.com/purpleidea/mgmt/event"
+	"github.com/purpleidea/mgmt/prometheus"
+	"github.com/purpleidea/mgmt/util"
 
 	errwrap "github.com/pkg/errors"
+	"golang.org/x/time/rate"
 )
 
 //go:generate stringer -type=ResState -output=resstate_stringer.go
@@ -41,22 +45,37 @@ type ResState int
 
 // Each ResState should be set properly in the relevant part of the resource.
 const (
-	ResStateNil ResState = iota
-	ResStateWatching
-	ResStateEvent // an event has happened, but we haven't poked yet
-	ResStateCheckApply
-	ResStatePoking
+	ResStateNil        ResState = iota
+	ResStateProcess             // we're in process, but we haven't done much yet
+	ResStateCheckApply          // we're about to run CheckApply
+	ResStatePoking              // we're done CheckApply, and we're about to poke
 )
 
 const refreshPathToken = "refresh"
 
+// World is an interface to the rest of the different graph state. It allows
+// the GAPI to store state and exchange information throughout the cluster. It
+// is the interface each machine uses to communicate with the rest of the world.
+type World interface { // TODO: is there a better name for this interface?
+	ResExport([]Res) error
+	// FIXME: should this method take a "filter" data struct instead of many args?
+	ResCollect(hostnameFilter, kindFilter []string) ([]Res, error)
+
+	StrWatch(namespace string) chan error
+	StrGet(namespace string) (map[string]string, error)
+	StrSet(namespace, value string) error
+	StrDel(namespace string) error
+}
+
 // Data is the set of input values passed into the pgraph for the resources.
 type Data struct {
-	//Hostname string         // uuid for the host
+	Hostname string // uuid for the host
 	//Noop     bool
-	Converger converger.Converger
-	Prefix    string // the prefix to be used for the pgraph namespace
-	Debug     bool
+	Converger  converger.Converger
+	Prometheus *prometheus.Prometheus
+	World      World
+	Prefix     string // the prefix to be used for the pgraph namespace
+	Debug      bool
 	// NOTE: we can add more fields here if needed for the resources.
 }
 
@@ -91,8 +110,12 @@ type MetaParams struct {
 	// NOTE: there are separate Watch and CheckApply retry and delay values,
 	// but I've decided to use the same ones for both until there's a proper
 	// reason to want to do something differently for the Watch errors.
-	Retry int16  `yaml:"retry"` // metaparam, number of times to retry on error. -1 for infinite
-	Delay uint64 `yaml:"delay"` // metaparam, number of milliseconds to wait between retries
+	Retry int16      `yaml:"retry"` // metaparam, number of times to retry on error. -1 for infinite
+	Delay uint64     `yaml:"delay"` // metaparam, number of milliseconds to wait between retries
+	Poll  uint32     `yaml:"poll"`  // metaparam, number of seconds between poll intervals, 0 to watch
+	Limit rate.Limit `yaml:"limit"` // metaparam, number of events per second to allow through
+	Burst int        `yaml:"burst"` // metaparam, number of events to allow in a burst
+	Sema  []string   `yaml:"sema"`  // metaparam, list of semaphore ids (id | id:count)
 }
 
 // UnmarshalYAML is the custom unmarshal handler for the MetaParams struct. It
@@ -114,8 +137,12 @@ var DefaultMetaParams = MetaParams{
 	AutoEdge:  true,
 	AutoGroup: true,
 	Noop:      false,
-	Retry:     0, // TODO: is this a good default?
-	Delay:     0, // TODO: is this a good default?
+	Retry:     0,        // TODO: is this a good default?
+	Delay:     0,        // TODO: is this a good default?
+	Poll:      0,        // defaults to watching for events
+	Limit:     rate.Inf, // defaults to no limit
+	Burst:     0,        // no burst needed on an infinite rate // TODO: is this a good default?
+	//Sema:      []string{},
 }
 
 // The Base interface is everything that is common to all resources.
@@ -126,15 +153,21 @@ type Base interface {
 	SetKind(string)
 	Kind() string
 	Meta() *MetaParams
-	Events() chan event.Event
-	AssociateData(*Data)
-	IsWatching() bool
-	SetWatching(bool)
+	Events() chan *event.Event
+	Data() *Data
+	IsWorking() bool
+	IsQuiescing() bool
+	QuiesceGroup() *sync.WaitGroup
+	WaitGroup() *sync.WaitGroup
+	Setup()
+	Reset()
+	Converger() converger.Converger
+	ConvergerUIDs() (converger.UID, converger.UID, converger.UID)
 	GetState() ResState
 	SetState(ResState)
-	DoSend(chan event.Event, string) (bool, error)
-	SendEvent(event.EventName, bool, bool) bool
-	ReadEvent(*event.Event) (bool, bool)   // TODO: optional here?
+	Event() error
+	SendEvent(event.Kind, error) error
+	ReadEvent(*event.Event) (*error, bool)
 	Refresh() bool                         // is there a pending refresh to run?
 	SetRefresh(bool)                       // set the refresh state of this resource
 	SendRecv(Res) (map[string]bool, error) // send->recv data passing function
@@ -147,19 +180,31 @@ type Base interface {
 	GetGroup() []Res    // return everyone grouped inside me
 	SetGroup([]Res)
 	VarDir(string) (string, error)
+	Running() error           // notify the engine that Watch started
+	Started() <-chan struct{} // returns when the resource has started
+	Stopped() <-chan struct{} // returns when the resource has stopped
+	Starter(bool)
+	Poll() error // poll alternative to watching :(
+	ProcessChan() chan *event.Event
+	ProcessSync() *sync.WaitGroup
+	ProcessExit()
+	Prometheus() *prometheus.Prometheus
 }
 
 // Res is the minimum interface you need to implement to define a new resource.
 type Res interface {
-	Base // include everything from the Base interface
+	Base          // include everything from the Base interface
+	Default() Res // return a struct with sane defaults as a Res
+	Validate() error
 	Init() error
-	//Validate() error    // TODO: this might one day be added
-	GetUIDs() []ResUID            // most resources only return one
-	Watch(chan event.Event) error // send on channel to signal process() events
+	Close() error
+	UIDs() []ResUID // most resources only return one
+	Watch() error   // send on channel to signal process() events
 	CheckApply(apply bool) (checkOK bool, err error)
 	AutoEdges() AutoEdge
 	Compare(Res) bool
 	CollectPattern(string) // XXX: temporary until Res collection is more advanced
+	//UnmarshalYAML(unmarshal func(interface{}) error) error // optional
 }
 
 // BaseRes is the base struct that gets used in every resource.
@@ -168,20 +213,67 @@ type BaseRes struct {
 	MetaParams MetaParams       `yaml:"meta"` // struct of all the metaparams
 	Recv       map[string]*Send // mapping of key to receive on from value
 
-	kind      string
-	events    chan event.Event
+	kind   string
+	data   Data
+	state  ResState
+	prefix string // base prefix for this resource
+
+	eventsLock *sync.Mutex // locks around sending and closing of events channel
+	eventsDone bool
+	eventsChan chan *event.Event
+
+	processLock *sync.Mutex
+	processDone bool
+	processChan chan *event.Event
+	processSync *sync.WaitGroup
+
 	converger converger.Converger // converged tracking
-	prefix    string              // base prefix for this resource
-	debug     bool
-	state     ResState
-	watching  bool  // is Watch() loop running ?
-	isStateOK bool  // whether the state is okay based on events or not
+	cuid      converger.UID
+	wcuid     converger.UID
+	pcuid     converger.UID
+
+	started   chan struct{} // closed when worker is started/running
+	stopped   chan struct{} // closed when worker is stopped/exited
+	isStarted bool          // did the started chan already close?
+	starter   bool          // does this have indegree == 0 ? XXX: usually?
+
+	quiescing    bool // are we quiescing (pause or exit)
+	quiesceGroup *sync.WaitGroup
+	waitGroup    *sync.WaitGroup
+	working      bool // is the Worker() loop running ?
+	debug        bool
+	isStateOK    bool // whether the state is okay based on events or not
+
 	isGrouped bool  // am i contained within a group?
 	grouped   []Res // list of any grouped resources
-	refresh   bool  // does this resource have a refresh to run?
+
+	refresh bool // does this resource have a refresh to run?
 	//refreshState StatefulBool // TODO: future stateful bool
 }
 
+// UnmarshalYAML is the custom unmarshal handler for the BaseRes struct. It is
+// primarily useful for setting the defaults, in particular if meta is absent!
+// FIXME: how come we can't get this to work properly without dropping fields?
+//func (obj *BaseRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+//	DefaultBaseRes := BaseRes{
+//		// without specifying a default here, if we don't specify *any*
+//		// meta parameters in the yaml file, then the UnmarshalYAML for
+//		// the MetaParams struct won't run, and we won't get defaults!
+//		MetaParams: DefaultMetaParams, // force a default
+//	}
+
+//	type rawBaseRes BaseRes           // indirection to avoid infinite recursion
+//	raw := rawBaseRes(DefaultBaseRes) // convert; the defaults go here
+//	//raw := rawBaseRes{}
+
+//	if err := unmarshal(&raw); err != nil {
+//		return err
+//	}
+
+//	*obj = BaseRes(raw) // restore from indirection with type conversion!
+//	return nil
+//}
+
 // UIDExistsInUIDs wraps the IFF method when used with a list of UID's.
 func UIDExistsInUIDs(uid ResUID, uids []ResUID) bool {
 	for _, u := range uids {
@@ -223,18 +315,80 @@ func (obj *BaseUID) Reversed() bool {
 	return *obj.reversed
 }
 
+// Validate reports any problems with the struct definition.
+func (obj *BaseRes) Validate() error {
+	isInf := obj.Meta().Limit == rate.Inf || math.IsInf(float64(obj.Meta().Limit), 1)
+	if obj.Meta().Burst == 0 && !isInf { // blocked
+		return fmt.Errorf("Permanently limited (rate != Inf, burst: 0)")
+	}
+	return nil
+}
+
 // Init initializes structures like channels if created without New constructor.
 func (obj *BaseRes) Init() error {
-	if obj.kind == "" {
-		return fmt.Errorf("Resource did not set kind!")
+	if obj.debug {
+		log.Printf("%s[%s]: Init()", obj.Kind(), obj.GetName())
 	}
-	obj.events = make(chan event.Event) // unbuffered chan to avoid stale events
+	if obj.kind == "" {
+		return fmt.Errorf("resource did not set kind")
+	}
+
+	obj.cuid = obj.Converger().Register()
+	obj.wcuid = obj.Converger().Register() // get a cuid for the worker!
+	obj.pcuid = obj.Converger().Register() // get a cuid for the process
+
+	obj.processLock = &sync.Mutex{} // lock around processChan closing and sending
+	obj.processDone = false         // did we close processChan ?
+	obj.processChan = make(chan *event.Event)
+	obj.processSync = &sync.WaitGroup{}
+
+	obj.quiescing = false // no quiesce operation is happening at the moment
+	obj.quiesceGroup = &sync.WaitGroup{}
+
+	obj.waitGroup = &sync.WaitGroup{} // Init and Close must be 1-1 matched!
+	obj.waitGroup.Add(1)
+	obj.working = true // Worker method should now be running...
+
+	// FIXME: force a sane default until UnmarshalYAML on *BaseRes works...
+	if obj.Meta().Burst == 0 && obj.Meta().Limit == 0 { // blocked
+		obj.Meta().Limit = rate.Inf
+	}
+	if math.IsInf(float64(obj.Meta().Limit), 1) { // yaml `.inf` -> rate.Inf
+		obj.Meta().Limit = rate.Inf
+	}
+
 	//dir, err := obj.VarDir("")
 	//if err != nil {
-	//	return errwrap.Wrapf(err, "VarDir failed in Init()")
+	//	return errwrap.Wrapf(err, "the VarDir failed in Init()")
 	//}
 	// TODO: this StatefulBool implementation could be eventually swappable
 	//obj.refreshState = &DiskBool{Path: path.Join(dir, refreshPathToken)}
+
+	if err := obj.Prometheus().AddManagedResource(fmt.Sprintf("%v[%v]", obj.Kind(), obj.GetName()), obj.Kind()); err != nil {
+		return errwrap.Wrapf(err, "could not increase prometheus counter!")
+	}
+
+	return nil
+}
+
+// Close shuts down and performs any cleanup.
+func (obj *BaseRes) Close() error {
+	if obj.debug {
+		log.Printf("%s[%s]: Close()", obj.Kind(), obj.GetName())
+	}
+
+	obj.pcuid.Unregister()
+	obj.wcuid.Unregister()
+	obj.cuid.Unregister()
+
+	obj.working = false // Worker method should now be closing...
+	close(obj.stopped)
+	obj.waitGroup.Done()
+
+	if err := obj.Prometheus().RemoveManagedResource(fmt.Sprintf("%v[%v]", obj.Kind(), obj.GetName()), obj.kind); err != nil {
+		return errwrap.Wrapf(err, "could not decrease prometheus counter!")
+	}
+
 	return nil
 }
 
@@ -264,25 +418,61 @@ func (obj *BaseRes) Meta() *MetaParams {
 }
 
 // Events returns the channel of events to listen on.
-func (obj *BaseRes) Events() chan event.Event {
-	return obj.events
+func (obj *BaseRes) Events() chan *event.Event {
+	return obj.eventsChan
 }
 
-// AssociateData associates some data with the object in question.
-func (obj *BaseRes) AssociateData(data *Data) {
-	obj.converger = data.Converger
-	obj.prefix = data.Prefix
-	obj.debug = data.Debug
+// Data returns an associable handle to some data passed in to the resource.
+func (obj *BaseRes) Data() *Data {
+	return &obj.data
 }
 
-// IsWatching tells us if the Watch() function is running.
-func (obj *BaseRes) IsWatching() bool {
-	return obj.watching
+// IsWorking tells us if the Worker() function is running. Not thread safe.
+func (obj *BaseRes) IsWorking() bool {
+	return obj.working
 }
 
-// SetWatching stores the status of if the Watch() function is running.
-func (obj *BaseRes) SetWatching(b bool) {
-	obj.watching = b
+// IsQuiescing returns if there is a quiesce operation in progress. Pause and
+// exit both meet this criteria, and this tells some systems to wind down, such
+// as the event replay mechanism.
+func (obj *BaseRes) IsQuiescing() bool { return obj.quiescing }
+
+// QuiesceGroup returns the sync group associated with the quiesce operations.
+func (obj *BaseRes) QuiesceGroup() *sync.WaitGroup { return obj.quiesceGroup }
+
+// WaitGroup returns a sync.WaitGroup which is open when the resource is done.
+// This is more useful than a closed channel signal, since it can be re-used
+// safely without having to recreate it and worry about stale channel handles.
+func (obj *BaseRes) WaitGroup() *sync.WaitGroup { return obj.waitGroup }
+
+// Setup does some work which must happen before the Worker starts. It happens
+// once per Worker startup. It can happen in parallel with other Setup calls, so
+// add locks around any operation that's not thread-safe.
+func (obj *BaseRes) Setup() {
+	obj.started = make(chan struct{}) // closes when started
+	obj.stopped = make(chan struct{}) // closes when stopped
+
+	obj.eventsLock = &sync.Mutex{}
+	obj.eventsDone = false
+	obj.eventsChan = make(chan *event.Event) // unbuffered chan to avoid stale events
+}
+
+// Reset from Setup. These can get called for different vertices in parallel.
+func (obj *BaseRes) Reset() {
+	return
+}
+
+// Converger returns the converger object used by the system. It can be used to
+// register new convergers if needed.
+func (obj *BaseRes) Converger() converger.Converger {
+	return obj.data.Converger
+}
+
+// ConvergerUIDs returns the ConvergerUIDs for the resource. This is called by
+// the various methods that need one of these ConvergerUIDs. They are registered
+// by the Init method and unregistered on the resource Close.
+func (obj *BaseRes) ConvergerUIDs() (cuid, wcuid, pcuid converger.UID) {
+	return obj.cuid, obj.wcuid, obj.pcuid
 }
 
 // GetState returns the state of the resource.
@@ -308,6 +498,21 @@ func (obj *BaseRes) StateOK(b bool) {
 	obj.isStateOK = b
 }
 
+// ProcessChan returns the chan that resources send events to. Internal API!
+func (obj *BaseRes) ProcessChan() chan *event.Event { return obj.processChan }
+
+// ProcessSync returns the WaitGroup that blocks until the innerWorker closes.
+func (obj *BaseRes) ProcessSync() *sync.WaitGroup { return obj.processSync }
+
+// ProcessExit causes the innerWorker to close and waits until it does so.
+func (obj *BaseRes) ProcessExit() {
+	obj.processLock.Lock() // lock to avoid a send when closed!
+	obj.processDone = true
+	close(obj.processChan)
+	obj.processLock.Unlock()
+	obj.processSync.Wait()
+}
+
 // GroupCmp compares two resources and decides if they're suitable for grouping
 // You'll probably want to override this method when implementing a resource...
 func (obj *BaseRes) GroupCmp(res Res) bool {
@@ -317,10 +522,16 @@ func (obj *BaseRes) GroupCmp(res Res) bool {
 // GroupRes groups resource (arg) into self.
 func (obj *BaseRes) GroupRes(res Res) error {
 	if l := len(res.GetGroup()); l > 0 {
-		return fmt.Errorf("Res: %v already contains %d grouped resources!", res, l)
+		return fmt.Errorf("the %v resource already contains %d grouped resources", res, l)
 	}
 	if res.IsGrouped() {
-		return fmt.Errorf("Res: %v is already grouped!", res)
+		return fmt.Errorf("the %v resource is already grouped", res)
+	}
+
+	// merging two resources into one should yield the sum of their semas
+	if semas := res.Meta().Sema; len(semas) > 0 {
+		obj.Meta().Sema = append(obj.Meta().Sema, semas...)
+		obj.Meta().Sema = util.StrRemoveDuplicatesInList(obj.Meta().Sema)
 	}
 
 	obj.grouped = append(obj.grouped, res)
@@ -371,6 +582,33 @@ func (obj *BaseRes) Compare(res Res) bool {
 	if obj.Meta().Delay != res.Meta().Delay {
 		return false
 	}
+	if obj.Meta().Poll != res.Meta().Poll {
+		return false
+	}
+	if obj.Meta().Limit != res.Meta().Limit {
+		return false
+	}
+	if obj.Meta().Burst != res.Meta().Burst {
+		return false
+	}
+
+	// are the two slices the same?
+	cmpSlices := func(a, b []string) bool {
+		if len(a) != len(b) {
+			return false
+		}
+		sort.Strings(a)
+		sort.Strings(b)
+		for i := range a {
+			if a[i] != b[i] {
+				return false
+			}
+		}
+		return true
+	}
+	if !cmpSlices(obj.Meta().Sema, res.Meta().Sema) {
+		return false
+	}
 	return true
 }
 
@@ -385,52 +623,72 @@ func (obj *BaseRes) VarDir(extra string) (string, error) {
 	// Using extra adds additional dirs onto our namespace. An empty extra
 	// adds no additional directories.
 	if obj.prefix == "" {
-		return "", fmt.Errorf("VarDir prefix is empty!")
+		return "", fmt.Errorf("the VarDir prefix is empty")
 	}
 	if obj.Kind() == "" {
-		return "", fmt.Errorf("VarDir kind is empty!")
+		return "", fmt.Errorf("the VarDir kind is empty")
 	}
 	if obj.GetName() == "" {
-		return "", fmt.Errorf("VarDir name is empty!")
+		return "", fmt.Errorf("the VarDir name is empty")
 	}
 
 	// FIXME: is obj.GetName() sufficiently unique to use as a UID here?
 	uid := obj.GetName()
 	p := fmt.Sprintf("%s/", path.Join(obj.prefix, obj.Kind(), uid, extra))
 	if err := os.MkdirAll(p, 0770); err != nil {
-		return "", errwrap.Wrapf(err, "Can't create prefix for %s[%s]", obj.Kind(), obj.GetName())
+		return "", errwrap.Wrapf(err, "can't create prefix for %s[%s]", obj.Kind(), obj.GetName())
 	}
 	return p, nil
 }
 
-// ResToB64 encodes a resource to a base64 encoded string (after serialization)
-func ResToB64(res Res) (string, error) {
-	b := bytes.Buffer{}
-	e := gob.NewEncoder(&b)
-	err := e.Encode(&res) // pass with &
-	if err != nil {
-		return "", fmt.Errorf("Gob failed to encode: %v", err)
+// Started returns a channel that closes when the resource has started up.
+func (obj *BaseRes) Started() <-chan struct{} { return obj.started }
+
+// Stopped returns a channel that closes when the worker has finished running.
+func (obj *BaseRes) Stopped() <-chan struct{} { return obj.stopped }
+
+// Starter sets the starter bool. This defines if a vertex has an indegree of 0.
+// If we have an indegree of 0, we'll need to be a poke initiator in the graph.
+func (obj *BaseRes) Starter(b bool) { obj.starter = b }
+
+// Poll is the watch replacement for when we want to poll, which outputs events.
+func (obj *BaseRes) Poll() error {
+	cuid, _, _ := obj.ConvergerUIDs() // get the converger uid used to report status
+
+	// create a time.Ticker for the given interval
+	ticker := time.NewTicker(time.Duration(obj.Meta().Poll) * time.Second)
+	defer ticker.Stop()
+
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+	cuid.SetConverged(false) // quickly stop any converge due to Running()
+
+	var send = false
+	var exit *error
+	for {
+		select {
+		case <-ticker.C: // received the timer event
+			log.Printf("%s[%s]: polling...", obj.Kind(), obj.GetName())
+			send = true
+			obj.StateOK(false) // dirty
+
+		case event := <-obj.Events():
+			cuid.ResetTimer() // important
+			if exit, send = obj.ReadEvent(event); exit != nil {
+				return *exit // exit
+			}
+		}
+
+		if send {
+			send = false
+			obj.Event()
+		}
 	}
-	return base64.StdEncoding.EncodeToString(b.Bytes()), nil
 }
 
-// B64ToRes decodes a resource from a base64 encoded string (after deserialization)
-func B64ToRes(str string) (Res, error) {
-	var output interface{}
-	bb, err := base64.StdEncoding.DecodeString(str)
-	if err != nil {
-		return nil, fmt.Errorf("Base64 failed to decode: %v", err)
-	}
-	b := bytes.NewBuffer(bb)
-	d := gob.NewDecoder(b)
-	err = d.Decode(&output) // pass with &
-	if err != nil {
-		return nil, fmt.Errorf("Gob failed to decode: %v", err)
-	}
-	res, ok := output.(Res)
-	if !ok {
-		return nil, fmt.Errorf("Output %v is not a Res", res)
-
-	}
-	return res, nil
+// Prometheus returns the prometheus instance.
+func (obj *BaseRes) Prometheus() *prometheus.Prometheus {
+	return obj.Data().Prometheus
 }

resources/resources_test.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -22,6 +22,7 @@ import (
 	"encoding/base64"
 	"encoding/gob"
 	"testing"
+	//"github.com/purpleidea/mgmt/event"
 )
 
 func TestMiscEncodeDecode1(t *testing.T) {
@@ -119,55 +120,55 @@ func TestIFF(t *testing.T) {
 }
 
 func TestReadEvent(t *testing.T) {
-	res := FileRes{}
+	//res := FileRes{}
 
-	shouldExit := map[eventName]bool{
-		eventStart:    false,
-		eventPoke:     false,
-		eventBackPoke: false,
-		eventExit:     true,
-	}
-	shouldPoke := map[eventName]bool{
-		eventStart:    true,
-		eventPoke:     true,
-		eventBackPoke: true,
-		eventExit:     false,
-	}
+	//shouldExit := map[event.Kind]bool{
+	//	event.EventStart:    false,
+	//	event.EventPoke:     false,
+	//	event.EventBackPoke: false,
+	//	event.EventExit:     true,
+	//}
+	//shouldPoke := map[event.Kind]bool{
+	//	event.EventStart:    true,
+	//	event.EventPoke:     true,
+	//	event.EventBackPoke: true,
+	//	event.EventExit:     false,
+	//}
 
-	for event := range shouldExit {
-		exit, poke := res.ReadEvent(&Event{Name: event})
-		if exit != shouldExit[event] {
-			t.Errorf("resource.ReadEvent returned wrong exit flag for a %v event (%v, should be %v)",
-				event, exit, shouldExit[event])
-		}
-		if poke != shouldPoke[event] {
-			t.Errorf("resource.ReadEvent returned wrong poke flag for a %v event (%v, should be %v)",
-				event, poke, shouldPoke[event])
-		}
-	}
+	//for ev := range shouldExit {
+	//	exit, poke := res.ReadEvent(&event.Event{Kind: ev})
+	//	if exit != shouldExit[ev] {
+	//		t.Errorf("resource.ReadEvent returned wrong exit flag for a %v event (%v, should be %v)",
+	//			ev, exit, shouldExit[ev])
+	//	}
+	//	if poke != shouldPoke[ev] {
+	//		t.Errorf("resource.ReadEvent returned wrong poke flag for a %v event (%v, should be %v)",
+	//			ev, poke, shouldPoke[ev])
+	//	}
+	//}
 
-	res.Init()
-	res.SetWatching(true)
+	//res.Init()
+	//res.SetWatching(true)
 
 	// test result when a pause event is followed by start
-	go res.SendEvent(eventStart, false, false)
-	exit, poke := res.ReadEvent(&Event{Name: eventPause})
-	if exit {
-		t.Error("resource.ReadEvent returned wrong exit flag for a pause+start event (true, should be false)")
-	}
-	if poke {
-		t.Error("resource.ReadEvent returned wrong poke flag for a pause+start event (true, should be false)")
-	}
+	//go res.SendEvent(event.EventStart, nil)
+	//exit, poke := res.ReadEvent(&event.Event{Kind: event.EventPause})
+	//if exit {
+	//	t.Error("resource.ReadEvent returned wrong exit flag for a pause+start event (true, should be false)")
+	//}
+	//if poke {
+	//	t.Error("resource.ReadEvent returned wrong poke flag for a pause+start event (true, should be false)")
+	//}
 
 	// test result when a pause event is followed by exit
-	go res.SendEvent(eventExit, false, false)
-	exit, poke = res.ReadEvent(&Event{Name: eventPause})
-	if !exit {
-		t.Error("resource.ReadEvent returned wrong exit flag for a pause+start event (false, should be true)")
-	}
-	if poke {
-		t.Error("resource.ReadEvent returned wrong poke flag for a pause+start event (true, should be false)")
-	}
+	//go res.SendEvent(event.EventExit, nil)
+	//exit, poke = res.ReadEvent(&event.Event{Kind: event.EventPause})
+	//if !exit {
+	//	t.Error("resource.ReadEvent returned wrong exit flag for a pause+start event (false, should be true)")
+	//}
+	//if poke {
+	//	t.Error("resource.ReadEvent returned wrong poke flag for a pause+start event (true, should be false)")
+	//}
 
 	// TODO: create a wrapper API around log, so that Fatals can be mocked and tested
 }

resources/sendrecv.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -28,84 +28,142 @@ import (
 	errwrap "github.com/pkg/errors"
 )
 
-// SendEvent pushes an event into the message queue for a particular vertex
-func (obj *BaseRes) SendEvent(ev event.EventName, sync bool, activity bool) bool {
-	// TODO: isn't this race-y ?
-	if !obj.IsWatching() { // element has already exited
-		return false // if we don't return, we'll block on the send
-	}
-	if !sync {
-		obj.events <- event.Event{Name: ev, Resp: nil, Msg: "", Activity: activity}
-		return true
-	}
-
+// Event sends off an event, but doesn't block the incoming event queue.
+func (obj *BaseRes) Event() error {
 	resp := event.NewResp()
-	obj.events <- event.Event{Name: ev, Resp: resp, Msg: "", Activity: activity}
-	resp.ACKWait() // waits until true (nil) value
-	return true
+	obj.processLock.Lock()
+	if obj.processDone {
+		obj.processLock.Unlock()
+		return fmt.Errorf("processChan is already closed")
+	}
+	obj.quiesceGroup.Add(1)                                           // add to processChan queue count
+	obj.processChan <- &event.Event{Kind: event.EventNil, Resp: resp} // trigger process
+	obj.processLock.Unlock()
+	return resp.Wait()
 }
 
-// DoSend sends off an event, but doesn't block the incoming event queue.
-func (obj *BaseRes) DoSend(processChan chan event.Event, comment string) (exit bool, err error) {
+// SendEvent pushes an event into the message queue for a particular vertex.
+func (obj *BaseRes) SendEvent(ev event.Kind, err error) error {
+	if obj.debug {
+		if err == nil {
+			log.Printf("%s[%s]: SendEvent(%+v)", obj.Kind(), obj.GetName(), ev)
+		} else {
+			log.Printf("%s[%s]: SendEvent(%+v): %v", obj.Kind(), obj.GetName(), ev, err)
+		}
+	}
 	resp := event.NewResp()
-	processChan <- event.Event{Name: event.EventNil, Resp: resp, Activity: false, Msg: comment} // trigger process
-	e := resp.Wait()
-	return false, e // XXX: at the moment, we don't use the exit bool.
+	obj.eventsLock.Lock()
+	if obj.eventsDone {
+		obj.eventsLock.Unlock()
+		return fmt.Errorf("eventsChan is already closed")
+	}
+	obj.eventsChan <- &event.Event{Kind: ev, Resp: resp, Err: err}
+	if ev == event.EventExit {
+		obj.eventsDone = true
+		close(obj.eventsChan) // this is where we properly close this channel!
+	}
+	obj.eventsLock.Unlock()
+	resp.ACKWait() // waits until true (nil) value
+	return nil
 }
 
 // ReadEvent processes events when a select gets one, and handles the pause
 // code too! The return values specify if we should exit and poke respectively.
-func (obj *BaseRes) ReadEvent(ev *event.Event) (exit, send bool) {
-	ev.ACK()
-	var poke bool
-	// ensure that a CheckApply runs by sending with a dirty state...
-	if ev.GetActivity() { // if previous node did work, and we were notified...
-		obj.StateOK(false) // dirty
-		poke = true        // poke!
-		// XXX: this should be elsewhere in case Watch isn't used (eg: Polling instead...)
-		// XXX: unless this is used in our "fallback" polling implementation???
-		//obj.SetRefresh(true) // TODO: is this redundant?
-	}
+func (obj *BaseRes) ReadEvent(ev *event.Event) (exit *error, send bool) {
+	//ev.ACK()
+	err := ev.Error()
 
-	switch ev.Name {
+	switch ev.Kind {
 	case event.EventStart:
-		send = true || poke
-		return
+		ev.ACK()
+		return nil, true
 
 	case event.EventPoke:
-		send = true || poke
-		return
+		ev.ACK()
+		return nil, true
 
 	case event.EventBackPoke:
-		send = true || poke
-		return // forward poking in response to a back poke!
+		ev.ACK()
+		return nil, true // forward poking in response to a back poke!
 
 	case event.EventExit:
+		obj.quiescing = true
+		obj.quiesceGroup.Wait()
+		obj.quiescing = false // for symmetry
+		ev.ACK()
 		// FIXME: what do we do if we have a pending refresh (poke) and an exit?
-		return true, false
+		return &err, false
 
 	case event.EventPause:
-		// wait for next event to continue
-		select {
-		case e, ok := <-obj.Events():
-			if !ok { // shutdown
-				return true, false
-			}
-			e.ACK()
-			if e.Name == event.EventExit {
-				return true, false
-			} else if e.Name == event.EventStart { // eventContinue
-				return false, false // don't poke on unpause!
-			} else {
+		obj.quiescing = true // set the quiesce flag to avoid event replays
+		obj.quiesceGroup.Wait()
+		obj.quiescing = false // reset
+		ev.ACK()
+
+		// wait for next event to continue, but discard any backpoking!
+		for {
+			// Consider a graph (V2->V3). If while paused, we add a
+			// new resource (V1->V2), when we unpause, V3 will run,
+			// and then V2 followed by V1 (reverse topo sort) which
+			// can cause V2 to BackPoke to V1 (since V1 needs to go
+			// first) which can panic if V1 is not running yet! The
+			// solution is to ignore the BackPoke because once that
+			// V1 vertex gets running, it will then send off a poke
+			// to V2 that it did without the need for the BackPoke!
+			select {
+			case e, ok := <-obj.Events():
+				if !ok { // shutdown
+					err := error(nil)
+					return &err, false
+				}
+				//obj.quiescing = true
+				//obj.quiesceGroup.Wait() // unnecessary, but symmetrically correct
+				//obj.quiescing = false
+				e.ACK()
+				err := e.Error()
+				if e.Kind == event.EventExit {
+					return &err, false
+				} else if e.Kind == event.EventStart { // eventContinue
+					return nil, false // don't poke on unpause!
+				} else if e.Kind == event.EventBackPoke {
+					continue // silently discard this event while paused
+				}
 				// if we get a poke event here, it's a bug!
-				log.Fatalf("%s[%s]: Unknown event: %v, while paused!", obj.Kind(), obj.GetName(), e)
+				err = fmt.Errorf("%s[%s]: unknown event: %v, while paused", obj.Kind(), obj.GetName(), e)
+				panic(err) // TODO: return a special sentinel instead?
+				//return &err, false
 			}
 		}
-
-	default:
-		log.Fatal("Unknown event: ", ev)
 	}
-	return true, false // required to keep the stupid go compiler happy
+	err = fmt.Errorf("unknown event: %v", ev)
+	panic(err) // TODO: return a special sentinel instead?
+	//return &err, false
+}
+
+// Running is called by the Watch method of the resource once it has started up.
+// This signals to the engine to kick off the initial CheckApply resource check.
+func (obj *BaseRes) Running() error {
+	// TODO: If a non-polling resource wants to use the converger, then it
+	// should probably tell Running (via an arg) to not do this. Currently
+	// it's a very unlikely race that could cause an early converge if the
+	// converge timeout is very short ( ~ 1s) and the Watch method doesn't
+	// immediately SetConverged(false) to stop possible early termination.
+	if obj.Meta().Poll == 0 { // if not polling, unblock this...
+		cuid, _, _ := obj.ConvergerUIDs()
+		cuid.SetConverged(true) // a reasonable initial assumption
+	}
+
+	obj.StateOK(false)  // assume we're initially dirty
+	if !obj.isStarted { // this avoids a double close when/if watch retries
+		obj.isStarted = true
+		close(obj.started) // send started signal
+	}
+
+	var err error
+	if obj.starter { // vertices of indegree == 0 should send initial pokes
+		err = obj.Event() // trigger a CheckApply
+	}
+	return err // bubble up any possible error (or nil)
 }
 
 // Send points to a value that a resource will send.
@@ -147,7 +205,7 @@ func (obj *BaseRes) SendRecv(res Res) (map[string]bool, error) {
 
 		// i think we probably want the same kind, at least for now...
 		if kind1 != kind2 {
-			e := fmt.Errorf("Kind mismatch between %s[%s]: %s and %s[%s]: %s", v.Res.Kind(), v.Res.GetName(), kind1, obj.Kind(), obj.GetName(), kind2)
+			e := fmt.Errorf("kind mismatch between %s[%s]: %s and %s[%s]: %s", v.Res.Kind(), v.Res.GetName(), kind1, obj.Kind(), obj.GetName(), kind2)
 			err = multierr.Append(err, e) // list of errors
 			continue
 		}
@@ -155,21 +213,21 @@ func (obj *BaseRes) SendRecv(res Res) (map[string]bool, error) {
 		// if the types don't match, we can't use send->recv
 		// TODO: do we want to relax this for string -> *string ?
 		if e := TypeCmp(value1, value2); e != nil {
-			e := errwrap.Wrapf(e, "Type mismatch between %s[%s] and %s[%s]", v.Res.Kind(), v.Res.GetName(), obj.Kind(), obj.GetName())
+			e := errwrap.Wrapf(e, "type mismatch between %s[%s] and %s[%s]", v.Res.Kind(), v.Res.GetName(), obj.Kind(), obj.GetName())
 			err = multierr.Append(err, e) // list of errors
 			continue
 		}
 
 		// if we can't set, then well this is pointless!
 		if !value2.CanSet() {
-			e := fmt.Errorf("Can't set %s[%s].%s", obj.Kind(), obj.GetName(), k)
+			e := fmt.Errorf("can't set %s[%s].%s", obj.Kind(), obj.GetName(), k)
 			err = multierr.Append(err, e) // list of errors
 			continue
 		}
 
 		// if we can't interface, we can't compare...
 		if !value1.CanInterface() || !value2.CanInterface() {
-			e := fmt.Errorf("Can't interface %s[%s].%s", obj.Kind(), obj.GetName(), k)
+			e := fmt.Errorf("can't interface %s[%s].%s", obj.Kind(), obj.GetName(), k)
 			err = multierr.Append(err, e) // list of errors
 			continue
 		}
@@ -191,7 +249,7 @@ func (obj *BaseRes) SendRecv(res Res) (map[string]bool, error) {
 func TypeCmp(a, b reflect.Value) error {
 	ta, tb := a.Type(), b.Type()
 	if ta != tb {
-		return fmt.Errorf("Type mismatch: %s != %s", ta, tb)
+		return fmt.Errorf("type mismatch: %s != %s", ta, tb)
 	}
 	// NOTE: it seems we don't need to recurse into pointers to sub check!
 

resources/svc.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -23,9 +23,7 @@ import (
 	"encoding/gob"
 	"fmt"
 	"log"
-	"time"
 
-	"github.com/purpleidea/mgmt/event"
 	"github.com/purpleidea/mgmt/util"
 
 	systemd "github.com/coreos/go-systemd/dbus" // change namespace
@@ -45,69 +43,49 @@ type SvcRes struct {
 	Startup string `yaml:"startup"` // enabled, disabled, undefined
 }
 
-// NewSvcRes is a constructor for this resource. It also calls Init() for you.
-func NewSvcRes(name, state, startup string) (*SvcRes, error) {
-	obj := &SvcRes{
+// Default returns some sensible defaults for this resource.
+func (obj *SvcRes) Default() Res {
+	return &SvcRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		State:   state,
-		Startup: startup,
 	}
-	return obj, obj.Init()
-}
-
-// Init runs some startup code for this resource.
-func (obj *SvcRes) Init() error {
-	obj.BaseRes.kind = "Svc"
-	return obj.BaseRes.Init() // call base init, b/c we're overriding
 }
 
 // Validate checks if the resource data structure was populated correctly.
 func (obj *SvcRes) Validate() error {
 	if obj.State != "running" && obj.State != "stopped" && obj.State != "" {
-		return fmt.Errorf("State must be either `running` or `stopped` or undefined.")
+		return fmt.Errorf("state must be either `running` or `stopped` or undefined")
 	}
 	if obj.Startup != "enabled" && obj.Startup != "disabled" && obj.Startup != "" {
-		return fmt.Errorf("Startup must be either `enabled` or `disabled` or undefined.")
+		return fmt.Errorf("startup must be either `enabled` or `disabled` or undefined")
 	}
-	return nil
+	return obj.BaseRes.Validate()
+}
+
+// Init runs some startup code for this resource.
+func (obj *SvcRes) Init() error {
+	obj.BaseRes.kind = "svc"
+	return obj.BaseRes.Init() // call base init, b/c we're overriding
 }
 
 // Watch is the primary listener for this resource and it outputs events.
-func (obj *SvcRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
-	}
-
+func (obj *SvcRes) Watch() error {
 	// obj.Name: svc name
 	if !systemdUtil.IsRunningSystemd() {
-		return fmt.Errorf("Systemd is not running.")
+		return fmt.Errorf("systemd is not running")
 	}
 
 	conn, err := systemd.NewSystemdConnection() // needs root access
 	if err != nil {
-		return errwrap.Wrapf(err, "Failed to connect to systemd")
+		return errwrap.Wrapf(err, "failed to connect to systemd")
 	}
 	defer conn.Close()
 
 	// if we share the bus with others, we will get each others messages!!
 	bus, err := util.SystemBusPrivateUsable() // don't share the bus connection!
 	if err != nil {
-		return errwrap.Wrapf(err, "Failed to connect to bus")
+		return errwrap.Wrapf(err, "failed to connect to bus")
 	}
 
 	// XXX: will this detect new units?
@@ -116,9 +94,14 @@ func (obj *SvcRes) Watch(processChan chan event.Event) error {
 	buschan := make(chan *dbus.Signal, 10)
 	bus.Signal(buschan)
 
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
 	var svc = fmt.Sprintf("%s.service", obj.Name) // systemd name
 	var send = false                              // send event?
-	var exit = false
+	var exit *error
 	var invalid = false              // does the svc exist or not?
 	var previous bool                // previous invalid value
 	set := conn.NewSubscriptionSet() // no error should be returned
@@ -159,27 +142,15 @@ func (obj *SvcRes) Watch(processChan chan event.Event) error {
 				set.Remove(svc) // no return value should ever occur
 			}
 
-			obj.SetState(ResStateWatching) // reset
 			select {
 			case <-buschan: // XXX: wait for new units event to unstick
-				cuid.SetConverged(false)
 				// loop so that we can see the changed invalid signal
 				log.Printf("Svc[%s]->DaemonReload()", svc)
 
 			case event := <-obj.Events():
-				cuid.SetConverged(false)
-				if exit, send = obj.ReadEvent(&event); exit {
-					return nil // exit
+				if exit, send = obj.ReadEvent(event); exit != nil {
+					return *exit // exit
 				}
-
-			case <-cuid.ConvergedTimer():
-				cuid.SetConverged(true) // converged!
-				continue
-
-			case <-Startup(startup):
-				cuid.SetConverged(false)
-				send = true
-				obj.StateOK(false) // dirty
 			}
 		} else {
 			if !activeSet {
@@ -188,7 +159,6 @@ func (obj *SvcRes) Watch(processChan chan event.Event) error {
 			}
 
 			log.Printf("Watching: %s", svc) // attempting to watch...
-			obj.SetState(ResStateWatching)  // reset
 			select {
 			case event := <-subChannel:
 
@@ -204,6 +174,8 @@ func (obj *SvcRes) Watch(processChan chan event.Event) error {
 						log.Printf("Svc[%s]->Stopped", svc)
 					case "reloading":
 						log.Printf("Svc[%s]->Reloading", svc)
+					case "failed":
+						log.Printf("Svc[%s]->Failed", svc)
 					default:
 						log.Fatalf("Unknown svc state: %s", event[svc].ActiveState)
 					}
@@ -215,32 +187,18 @@ func (obj *SvcRes) Watch(processChan chan event.Event) error {
 				obj.StateOK(false) // dirty
 
 			case err := <-subErrors:
-				cuid.SetConverged(false)
-				return errwrap.Wrapf(err, "Unknown %s[%s] error", obj.Kind(), obj.GetName())
+				return errwrap.Wrapf(err, "unknown %s[%s] error", obj.Kind(), obj.GetName())
 
 			case event := <-obj.Events():
-				cuid.SetConverged(false)
-				if exit, send = obj.ReadEvent(&event); exit {
-					return nil // exit
+				if exit, send = obj.ReadEvent(event); exit != nil {
+					return *exit // exit
 				}
-
-			case <-cuid.ConvergedTimer():
-				cuid.SetConverged(true) // converged!
-				continue
-
-			case <-Startup(startup):
-				cuid.SetConverged(false)
-				send = true
-				obj.StateOK(false) // dirty
 			}
 		}
 
 		if send {
-			startup = true // startup finished
 			send = false
-			if exit, err := obj.DoSend(processChan, ""); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -249,12 +207,12 @@ func (obj *SvcRes) Watch(processChan chan event.Event) error {
 // input is true. It returns error info and if the state check passed or not.
 func (obj *SvcRes) CheckApply(apply bool) (checkOK bool, err error) {
 	if !systemdUtil.IsRunningSystemd() {
-		return false, fmt.Errorf("Systemd is not running.")
+		return false, fmt.Errorf("systemd is not running")
 	}
 
 	conn, err := systemd.NewSystemdConnection() // needs root access
 	if err != nil {
-		return false, errwrap.Wrapf(err, "Failed to connect to systemd")
+		return false, errwrap.Wrapf(err, "failed to connect to systemd")
 	}
 	defer conn.Close()
 
@@ -262,13 +220,13 @@ func (obj *SvcRes) CheckApply(apply bool) (checkOK bool, err error) {
 
 	loadstate, err := conn.GetUnitProperty(svc, "LoadState")
 	if err != nil {
-		return false, errwrap.Wrapf(err, "Failed to get load state")
+		return false, errwrap.Wrapf(err, "failed to get load state")
 	}
 
 	// NOTE: we have to compare variants with other variants, they are really strings...
 	var notFound = (loadstate.Value == dbus.MakeVariant("not-found"))
 	if notFound {
-		return false, errwrap.Wrapf(err, "Failed to find svc: %s", svc)
+		return false, errwrap.Wrapf(err, "failed to find svc: %s", svc)
 	}
 
 	// XXX: check svc "enabled at boot" or not status...
@@ -276,7 +234,7 @@ func (obj *SvcRes) CheckApply(apply bool) (checkOK bool, err error) {
 	//conn.GetUnitProperties(svc)
 	activestate, err := conn.GetUnitProperty(svc, "ActiveState")
 	if err != nil {
-		return false, errwrap.Wrapf(err, "Failed to get active state")
+		return false, errwrap.Wrapf(err, "failed to get active state")
 	}
 
 	var running = (activestate.Value == dbus.MakeVariant("active"))
@@ -304,7 +262,7 @@ func (obj *SvcRes) CheckApply(apply bool) (checkOK bool, err error) {
 	}
 
 	if err != nil {
-		return false, errwrap.Wrapf(err, "Unable to change startup status")
+		return false, errwrap.Wrapf(err, "unable to change startup status")
 	}
 
 	// XXX: do we need to use a buffered channel here?
@@ -313,7 +271,7 @@ func (obj *SvcRes) CheckApply(apply bool) (checkOK bool, err error) {
 	if obj.State == "running" {
 		_, err = conn.StartUnit(svc, "fail", result)
 		if err != nil {
-			return false, errwrap.Wrapf(err, "Failed to start unit")
+			return false, errwrap.Wrapf(err, "failed to start unit")
 		}
 		if refresh {
 			log.Printf("%s[%s]: Skipping reload, due to pending start", obj.Kind(), obj.GetName())
@@ -322,7 +280,7 @@ func (obj *SvcRes) CheckApply(apply bool) (checkOK bool, err error) {
 	} else if obj.State == "stopped" {
 		_, err = conn.StopUnit(svc, "fail", result)
 		if err != nil {
-			return false, errwrap.Wrapf(err, "Failed to stop unit")
+			return false, errwrap.Wrapf(err, "failed to stop unit")
 		}
 		if refresh {
 			log.Printf("%s[%s]: Skipping reload, due to pending stop", obj.Kind(), obj.GetName())
@@ -332,10 +290,10 @@ func (obj *SvcRes) CheckApply(apply bool) (checkOK bool, err error) {
 
 	status := <-result
 	if &status == nil {
-		return false, fmt.Errorf("Systemd service action result is nil")
+		return false, fmt.Errorf("systemd service action result is nil")
 	}
 	if status != "done" {
-		return false, fmt.Errorf("Unknown systemd return string: %v", status)
+		return false, fmt.Errorf("unknown systemd return string: %v", status)
 	}
 
 	if refresh { // we need to reload the service
@@ -377,7 +335,7 @@ type SvcResAutoEdges struct {
 // Next returns the next automatic edge.
 func (obj *SvcResAutoEdges) Next() []ResUID {
 	if obj.found {
-		log.Fatal("Shouldn't be called anymore!")
+		log.Fatal("shouldn't be called anymore!")
 	}
 	if len(obj.data) == 0 { // check length for rare scenarios
 		return nil
@@ -397,7 +355,7 @@ func (obj *SvcResAutoEdges) Test(input []bool) bool {
 		return false
 	}
 	if len(input) != 1 { // in case we get given bad data
-		log.Fatal("Expecting a single value!")
+		log.Fatal("expecting a single value")
 	}
 	if input[0] { // if a match is found, we're done!
 		obj.found = true // no more to find!
@@ -431,9 +389,9 @@ func (obj *SvcRes) AutoEdges() AutoEdge {
 	}
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *SvcRes) GetUIDs() []ResUID {
+func (obj *SvcRes) UIDs() []ResUID {
 	x := &SvcUID{
 		BaseUID: BaseUID{name: obj.GetName(), kind: obj.Kind()},
 		name:    obj.Name, // svc name
@@ -476,3 +434,23 @@ func (obj *SvcRes) Compare(res Res) bool {
 	}
 	return true
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *SvcRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes SvcRes // indirection to avoid infinite recursion
+
+	def := obj.Default()     // get the default
+	res, ok := def.(*SvcRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to SvcRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = SvcRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/timer.go

@@ -1,5 +1,5 @@
 // Mgmt
-// Copyright (C) 2013-2016+ James Shubin and the project contributors
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
 // Written by James Shubin <james@shubin.ca> and the project contributors
 //
 // This program is free software: you can redistribute it and/or modify
@@ -19,10 +19,9 @@ package resources
 
 import (
 	"encoding/gob"
+	"fmt"
 	"log"
 	"time"
-
-	"github.com/purpleidea/mgmt/event"
 )
 
 func init() {
@@ -32,7 +31,7 @@ func init() {
 // TimerRes is a timer resource for time based events.
 type TimerRes struct {
 	BaseRes  `yaml:",inline"`
-	Interval int `yaml:"interval"` // Interval : Interval between runs
+	Interval uint32 `yaml:"interval"` // Interval : Interval between runs
 
 	ticker *time.Ticker
 }
@@ -43,87 +42,59 @@ type TimerUID struct {
 	name string
 }
 
-// NewTimerRes is a constructor for this resource. It also calls Init() for you.
-func NewTimerRes(name string, interval int) (*TimerRes, error) {
-	obj := &TimerRes{
+// Default returns some sensible defaults for this resource.
+func (obj *TimerRes) Default() Res {
+	return &TimerRes{
 		BaseRes: BaseRes{
-			Name: name,
+			MetaParams: DefaultMetaParams, // force a default
 		},
-		Interval: interval,
 	}
-	return obj, obj.Init()
+}
+
+// Validate the params that are passed to TimerRes.
+func (obj *TimerRes) Validate() error {
+	return obj.BaseRes.Validate()
 }
 
 // Init runs some startup code for this resource.
 func (obj *TimerRes) Init() error {
-	obj.BaseRes.kind = "Timer"
+	obj.BaseRes.kind = "timer"
 	return obj.BaseRes.Init() // call base init, b/c we're overrriding
 }
 
-// Validate the params that are passed to TimerRes
-// Currently we are getting only an interval in seconds
-// which gets validated by go compiler
-func (obj *TimerRes) Validate() error {
-	return nil
-}
-
 // newTicker creates a new ticker
 func (obj *TimerRes) newTicker() *time.Ticker {
 	return time.NewTicker(time.Duration(obj.Interval) * time.Second)
 }
 
 // Watch is the primary listener for this resource and it outputs events.
-func (obj *TimerRes) Watch(processChan chan event.Event) error {
-	if obj.IsWatching() {
-		return nil
-	}
-	obj.SetWatching(true)
-	defer obj.SetWatching(false)
-	cuid := obj.converger.Register()
-	defer cuid.Unregister()
-
-	var startup bool
-	Startup := func(block bool) <-chan time.Time {
-		if block {
-			return nil // blocks forever
-			//return make(chan time.Time) // blocks forever
-		}
-		return time.After(time.Duration(500) * time.Millisecond) // 1/2 the resolution of converged timeout
-	}
-
+func (obj *TimerRes) Watch() error {
 	// create a time.Ticker for the given interval
 	obj.ticker = obj.newTicker()
 	defer obj.ticker.Stop()
 
+	// notify engine that we're running
+	if err := obj.Running(); err != nil {
+		return err // bubble up a NACK...
+	}
+
 	var send = false
 
 	for {
-		obj.SetState(ResStateWatching)
 		select {
 		case <-obj.ticker.C: // received the timer event
 			send = true
 			log.Printf("%s[%s]: received tick", obj.Kind(), obj.GetName())
 
 		case event := <-obj.Events():
-			cuid.SetConverged(false)
-			if exit, _ := obj.ReadEvent(&event); exit {
-				return nil
+			if exit, _ := obj.ReadEvent(event); exit != nil {
+				return *exit // exit
 			}
-
-		case <-cuid.ConvergedTimer():
-			cuid.SetConverged(true)
-			continue
-
-		case <-Startup(startup):
-			cuid.SetConverged(false)
-			send = true
 		}
+
 		if send {
-			startup = true // startup finished
 			send = false
-			if exit, err := obj.DoSend(processChan, "timer ticked"); exit || err != nil {
-				return err // we exit or bubble up a NACK...
-			}
+			obj.Event()
 		}
 	}
 }
@@ -144,9 +115,9 @@ func (obj *TimerRes) CheckApply(apply bool) (bool, error) {
 	return false, nil
 }
 
-// GetUIDs includes all params to make a unique identification of this object.
+// UIDs includes all params to make a unique identification of this object.
 // Most resources only return one, although some resources can return multiple.
-func (obj *TimerRes) GetUIDs() []ResUID {
+func (obj *TimerRes) UIDs() []ResUID {
 	x := &TimerUID{
 		BaseUID: BaseUID{
 			name: obj.GetName(),
@@ -181,3 +152,23 @@ func (obj *TimerRes) Compare(res Res) bool {
 	}
 	return true
 }
+
+// UnmarshalYAML is the custom unmarshal handler for this struct.
+// It is primarily useful for setting the defaults.
+func (obj *TimerRes) UnmarshalYAML(unmarshal func(interface{}) error) error {
+	type rawRes TimerRes // indirection to avoid infinite recursion
+
+	def := obj.Default()       // get the default
+	res, ok := def.(*TimerRes) // put in the right format
+	if !ok {
+		return fmt.Errorf("could not convert to TimerRes")
+	}
+	raw := rawRes(*res) // convert; the defaults go here
+
+	if err := unmarshal(&raw); err != nil {
+		return err
+	}
+
+	*obj = TimerRes(raw) // restore from indirection with type conversion!
+	return nil
+}

resources/util.go

@@ -0,0 +1,59 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+
+package resources
+
+import (
+	"bytes"
+	"encoding/base64"
+	"encoding/gob"
+	"fmt"
+
+	errwrap "github.com/pkg/errors"
+)
+
+// ResToB64 encodes a resource to a base64 encoded string (after serialization).
+func ResToB64(res Res) (string, error) {
+	b := bytes.Buffer{}
+	e := gob.NewEncoder(&b)
+	err := e.Encode(&res) // pass with &
+	if err != nil {
+		return "", errwrap.Wrapf(err, "gob failed to encode")
+	}
+	return base64.StdEncoding.EncodeToString(b.Bytes()), nil
+}
+
+// B64ToRes decodes a resource from a base64 encoded string (after deserialization).
+func B64ToRes(str string) (Res, error) {
+	var output interface{}
+	bb, err := base64.StdEncoding.DecodeString(str)
+	if err != nil {
+		return nil, errwrap.Wrapf(err, "base64 failed to decode")
+	}
+	b := bytes.NewBuffer(bb)
+	d := gob.NewDecoder(b)
+	err = d.Decode(&output) // pass with &
+	if err != nil {
+		return nil, errwrap.Wrapf(err, "gob failed to decode")
+	}
+	res, ok := output.(Res)
+	if !ok {
+		return nil, fmt.Errorf("Output %v is not a Res", res)
+
+	}
+	return res, nil
+}

resources/virt_disabled.go

@@ -0,0 +1,24 @@
+// Mgmt
+// Copyright (C) 2013-2017+ James Shubin and the project contributors
+// Written by James Shubin <james@shubin.ca> and the project contributors
+//
+// This program is free software: you can redistribute it and/or modify
+// it under the terms of the GNU Affero General Public License as published by
+// the Free Software Foundation, either version 3 of the License, or
+// (at your option) any later version.
+//
+// This program is distributed in the hope that it will be useful,
+// but WITHOUT ANY WARRANTY; without even the implied warranty of
+// MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
+// GNU Affero General Public License for more details.
+//
+// You should have received a copy of the GNU Affero General Public License
+// along with this program.  If not, see <http://www.gnu.org/licenses/>.
+// +build novirt
+
+package resources
+
+// VirtRes represents the fields of the Virt resource. Since this file is
+// only invoked with the tag "novirt", we do not need any fields here.
+type VirtRes struct {
+}

spec.in

@@ -13,18 +13,18 @@ Source0:	https://dl.fedoraproject.org/pub/alt/purpleidea/__PROGRAM__/SOURCES/__P
 Requires: graphviz
 
 # If go_compiler is not set to 1, there is no virtual provide. Use golang instead.
-BuildRequires:  %{?go_compiler:compiler(go-compiler)}%{!?go_compiler:golang}
+BuildRequires:	%{?go_compiler:compiler(go-compiler)}%{!?go_compiler:golang}
 BuildRequires:	golang-googlecode-tools-stringer
 BuildRequires:	git-core
 BuildRequires:	mercurial
 
-ExclusiveArch:  %{go_arches}
+ExclusiveArch:	%{go_arches}
 
 %description
 A next generation config management prototype!
 
 %prep
-%setup
+%setup -q
 
 %build
 # FIXME: in the future, these could be vendor-ed in

tag.sh

@@ -7,5 +7,7 @@ echo "Version $t is now tagged!"
 echo "Pushing $t to origin..."
 echo "Press ^C within 3s to abort."
 sleep 3s
-git tag $t
+echo "release: tag $t" | git tag --file=- --sign $t
 git push origin $t
+git diff --stat "$v" "$t"
+if which contrib.sh 2>/dev/null; then contrib.sh "$v"; fi

test.sh

@@ -21,13 +21,14 @@ run-test ./test/test-gofmt.sh
 run-test ./test/test-yamlfmt.sh
 run-test ./test/test-bashfmt.sh
 run-test ./test/test-headerfmt.sh
-run-test go test
+run-test ./test/test-commit-message.sh
 run-test ./test/test-govet.sh
+run-test ./test/test-gotest.sh
 
 # do these longer tests only when running on ci
 if env | grep -q -e '^TRAVIS=true$' -e '^JENKINS_URL=' -e '^BUILD_TAG=jenkins'; then
-	run-test go test -race
 	run-test ./test/test-shell.sh
+	run-test ./test/test-gotest.sh --race
 fi
 
 # FIXME: this now fails everywhere :(

test/shell/augeas-1.sh

@@ -0,0 +1,30 @@
+#!/bin/bash -e
+
+if env | grep -q -e '^TRAVIS=true$'; then
+	# inotify doesn't seem to work properly on travis
+	echo "Travis and Jenkins give wonky results here, skipping test!"
+	exit
+fi
+
+mkdir -p "${MGMT_TMPDIR}"
+> "${MGMT_TMPDIR}"sshd_config
+
+# run empty graph, with prometheus support
+timeout --kill-after=20s 15s ./mgmt run --tmp-prefix --yaml=augeas-1.yaml &
+pid=$!
+sleep 5s	# let it converge
+
+grep "X11Forwarding no" "${MGMT_TMPDIR}"sshd_config
+
+sed -i "s/no/yes/" "${MGMT_TMPDIR}"sshd_config
+
+grep "X11Forwarding yes" "${MGMT_TMPDIR}"sshd_config
+
+sleep 3	# Augeas is slow
+
+grep "X11Forwarding no" "${MGMT_TMPDIR}"sshd_config
+
+
+killall -SIGINT mgmt	# send ^C to exit mgmt
+wait $pid	# get exit status
+exit $?