etcd: Rewrite embed etcd implementation

This is a giant cleanup of the etcd code. The earlier version was
written when I was less experienced with golang.

This is still not perfect, and does contain some races, but at least
it's a decent base to start from. The automatic elastic clustering
should be considered an experimental feature. If you need a more
battle-tested cluster, then you should manage etcd manually and point
mgmt at your existing cluster.
This commit is contained in:
James Shubin
2018-05-05 17:35:08 -04:00
parent fb275d9537
commit a5842a41b2
56 changed files with 5459 additions and 2654 deletions

View File

@@ -10,7 +10,7 @@ if ! command -v etcdctl >/dev/null; then
exit 0
fi
mkdir /tmp/mgmt/{A..E}
#mkdir /tmp/mgmt/{A..E}
# kill servers on error/exit
trap 'pkill -9 mgmt' EXIT
@@ -22,7 +22,7 @@ $TIMEOUT "$MGMT" run --hostname h3 --tmp-prefix --no-pgp --seeds http://127.0.0.
# wait for everything to converge
sleep 30s
ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2379 put /_mgmt/idealClusterSize 3
ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2379 put /_mgmt/chooser/dynamicsize/idealclustersize 3
$TIMEOUT "$MGMT" run --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 empty &
$TIMEOUT "$MGMT" run --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 empty &
@@ -32,7 +32,7 @@ sleep 30s
test "$(ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2379 member list | wc -l)" -eq 3
ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2381 put /_mgmt/idealClusterSize 5
ETCDCTL_API=3 etcdctl --endpoints 127.0.0.1:2379 put /_mgmt/chooser/dynamicsize/idealclustersize 5
# wait for everything to converge
sleep 30s

View File

@@ -0,0 +1,27 @@
#!/bin/bash
. "$(dirname "$0")/../util.sh"
# run empty graphs, we're just testing etcd clustering
$TIMEOUT "$MGMT" run --no-pgp --tmp-prefix empty &
pid1=$!
sleep 15s # let it startup
# run a second one that should conflict because a server is already running...
$TIMEOUT "$MGMT" run --no-pgp --tmp-prefix empty &
pid2=$!
wait $pid2
e=$?
if [ $e -eq 0 ]; then
echo "second mgmt exited successfully when error was expected"
exit 1
fi
if [ $e -ne 1 ]; then
echo "second mgmt exited with unexpected error of $e"
exit $e
fi
$(kill -SIGINT $pid1)& # send ^C to exit 1st mgmt
wait $pid1 # get exit status
# if pid1 exits because of a timeout, then it blocked, and this is a bug!
exit $?

View File

@@ -0,0 +1,35 @@
#!/bin/bash -e
. "$(dirname "$0")/../util.sh"
# run empty graphs, we're just testing etcd clustering
$TIMEOUT "$MGMT" run --hostname h1 --tmp-prefix empty &
pid1=$!
sleep 15s # let it startup
$TIMEOUT "$MGMT" run --hostname h2 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2381 --server-urls http://127.0.0.1:2382 --tmp-prefix empty &
pid2=$!
sleep 15s
$TIMEOUT "$MGMT" run --hostname h3 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2383 --server-urls http://127.0.0.1:2384 --tmp-prefix empty &
pid3=$!
sleep 15s
$(sleep 15s && kill -SIGINT $pid1)& # send ^C to exit 1st mgmt (reversed!)
wait $pid1
e=$?
if [ $e -ne 0 ]; then
exit $e
fi
$(sleep 15s && kill -SIGINT $pid2)& # send ^C to exit 2nd mgmt
wait $pid2
e=$?
if [ $e -ne 0 ]; then
exit $e
fi
$(sleep 15s && kill -SIGINT $pid3)& # send ^C to exit 3rd mgmt (reversed!)
wait $pid3 # get exit status
# if pid3 exits because of a timeout, then it blocked, and this is a bug!
exit $?

View File

@@ -0,0 +1,24 @@
#!/bin/bash -e
. "$(dirname "$0")/../util.sh"
# run empty graphs, we're just testing etcd clustering
$TIMEOUT "$MGMT" run --hostname h1 --tmp-prefix empty &
pid1=$!
sleep 15s # let it startup
$TIMEOUT "$MGMT" run --hostname h2 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2381 --server-urls http://127.0.0.1:2382 --tmp-prefix empty &
pid2=$!
sleep 15s
$(sleep 15s && kill -SIGINT $pid1)& # send ^C to exit 1st mgmt! (reverse!)
wait $pid1
e=$?
if [ $e -ne 0 ]; then
exit $e
fi
$(sleep 15s && kill -SIGINT $pid2)& # send ^C to exit 2nd mgmt (reverse!)
wait $pid2 # get exit status
# if pid2 exits because of a timeout, then it blocked, and this is a bug!
exit $?

View File

@@ -5,18 +5,58 @@
set -o errexit
set -o pipefail
$TIMEOUT "$MGMT" run --hostname h1 --ideal-cluster-size 1 --tmp-prefix --no-pgp lang --lang exchange0.mcl &
$TIMEOUT "$MGMT" run --hostname h2 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2381 --server-urls http://127.0.0.1:2382 --tmp-prefix --no-pgp lang --lang exchange0.mcl &
$TIMEOUT "$MGMT" run --hostname h3 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2383 --server-urls http://127.0.0.1:2384 --tmp-prefix --no-pgp lang --lang exchange0.mcl &
$TIMEOUT "$MGMT" run --hostname h4 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2385 --server-urls http://127.0.0.1:2386 --tmp-prefix --no-pgp lang --lang exchange0.mcl &
$TIMEOUT "$MGMT" run --hostname h1 --tmp-prefix --no-pgp empty &
pid1=$!
sleep 10s
$TIMEOUT "$MGMT" run --hostname h2 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2381 --server-urls http://127.0.0.1:2382 --tmp-prefix --no-pgp empty &
pid2=$!
sleep 10s
$TIMEOUT "$MGMT" run --hostname h3 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2383 --server-urls http://127.0.0.1:2384 --tmp-prefix --no-pgp empty &
pid3=$!
sleep 10s
$TIMEOUT "$MGMT" run --hostname h4 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2385 --server-urls http://127.0.0.1:2386 --tmp-prefix --no-pgp empty &
pid4=$!
sleep 10s
$TIMEOUT "$MGMT" deploy --no-git --seeds http://127.0.0.1:2379 lang --lang exchange0.mcl
# kill servers on error/exit
trap 'pkill -9 mgmt' EXIT
#trap 'pkill -9 mgmt' EXIT
# wait for everything to converge
sleep 10s
sleep 15s
# debug
tail /tmp/mgmt/exchange-*
test "$(cat /tmp/mgmt/exchange-* | grep -c h1)" -eq 4
test "$(cat /tmp/mgmt/exchange-* | grep -c h2)" -eq 4
test "$(cat /tmp/mgmt/exchange-* | grep -c h3)" -eq 4
test "$(cat /tmp/mgmt/exchange-* | grep -c h4)" -eq 4
$(sleep 15s && kill -SIGINT $pid4)& # send ^C to exit mgmt...
wait $pid4
e=$?
if [ $e -ne 0 ]; then
exit $e
fi
$(sleep 15s && kill -SIGINT $pid3)& # send ^C to exit mgmt...
wait $pid3
e=$?
if [ $e -ne 0 ]; then
exit $e
fi
$(sleep 15s && kill -SIGINT $pid2)& # send ^C to exit mgmt...
wait $pid2
e=$?
if [ $e -ne 0 ]; then
exit $e
fi
$(sleep 15s && kill -SIGINT $pid1)& # send ^C to exit mgmt...
wait $pid1
e=$?
if [ $e -ne 0 ]; then
exit $e
fi

View File

@@ -1,9 +1,10 @@
# run this example with these commands
# watch -n 0.1 'tail *' # run this in /tmp/mgmt/
# time ./mgmt run --hostname h1 --ideal-cluster-size 1 --tmp-prefix --no-pgp lang --lang examples/lang/exchange0.mcl
# time ./mgmt run --hostname h2 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2381 --server-urls http://127.0.0.1:2382 --tmp-prefix --no-pgp lang --lang examples/lang/exchange0.mcl
# time ./mgmt run --hostname h3 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2383 --server-urls http://127.0.0.1:2384 --tmp-prefix --no-pgp lang --lang examples/lang/exchange0.mcl
# time ./mgmt run --hostname h4 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2385 --server-urls http://127.0.0.1:2386 --tmp-prefix --no-pgp lang --lang examples/lang/exchange0.mcl
# time ./mgmt run --hostname h1 --tmp-prefix --no-pgp empty
# time ./mgmt run --hostname h2 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2381 --server-urls http://127.0.0.1:2382 --tmp-prefix --no-pgp empty
# time ./mgmt run --hostname h3 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2383 --server-urls http://127.0.0.1:2384 --tmp-prefix --no-pgp empty
# time ./mgmt run --hostname h4 --seeds http://127.0.0.1:2379 --client-urls http://127.0.0.1:2385 --server-urls http://127.0.0.1:2386 --tmp-prefix --no-pgp empty
# time ./mgmt deploy --no-git --seeds http://127.0.0.1:2379 lang --lang examples/lang/exchange0.mcl
import "sys"
import "world"