This adds an initial implementation of a virt resource based on libvirt.
It is not complete and requires more testing. The initial skeleton was
written by nseps but was not merged. It was later cleaned up and merged
in its current form by purpleidea. Many thanks to nseps for getting this
going, and hopefully we'll get you contributing more in the future!
Felix pointed out that these ID's are unique, but not universally unique
across the cluster, which might be confusing to new programmers. As a
result, rename them all.
This splits the recursive watching bit of the file file resource into
it's own package. This also de-duplicates the configwatch code and puts
it into the same package. With these bits refactored, it was also easier
to clean up the error code in main.
This was a sneaky deadlock which wasn't 100% reproducible. It was pretty
tricky to find because the same deadlocked behaviour was seen due to the
regression in the fsnotify module.
The event system needs a bit of a cleaning, but this can happen later.
This was a major deadlock that hit the file resource. I didn't notice it
earlier because I was using an older version of fsnotify and I hadn't
done a go get -u to refresh it. I finally tracked this down, and will
vendor the repository until a fix upstream or a workaround downstream is
added.
The upstream issue is: https://github.com/fsnotify/fsnotify/issues/123
This makes this logically more separate! :) As an aside...
I really hate the way golang does dependencies and packages. Yes, some
people insist on nesting their code deep into a $GOPATH, which is fine
if you're a google dev and are forced to work this way, but annoying for
the rest of the world. Your code shouldn't need a git commit to switch
to a a different vcs host! Gah I hate this so much.
All resources can now set a retry limit (-1 for infinite) and a delay
between retries. This applies to both the CheckApply methods, and the
Watch methods as well. They each have their own separate counts, but use
the same input meta param, since I decided it wouldn't be useful to have
a separate watchRetry and watchDelay set of meta parameters.
In the process, we got rid of about 15 error cases which would normally
panic.
This patch required a slight overhaul of the Event system.
The previous commit is an earlier version of this patch which I decided
to leave in to "show my work" as I used to have to do in math class.
It's slightly more correct with the current event system, and this
version is less correct and has a few bugs, but that is because the
event system needs a massive overhaul, and once that's done this should
all work properly for the corner cases.
This was the initial cut of the retry and delay meta parameters.
Instead, I decided to move the delay action into the common space
outside of the Watch resource. This is more complicated in the short
term, but will be more beneficial in the long run as each resource won't
have to implement this part itself (even if it uses boiler plate).
This is the first version of this patch without this fix. I decided to
include it because I think it has more correct event processing.
