plugin prevents successful weave launch on reboot

[rade]

The plugin is started with a restart policy of always. Consequently, if you have a startup script that runs weave launch, such as our published systemd, it will fail the 2nd time round, since weave launch will report weaveplugin is already running; you can stop it with 'weave stop-plugin'. and exit with status 1.

[bboreham]

If there was a $WEAVEPLUGIN_DOCKER_ARGS, then folks who didn't want the restart policy could put --restart=none there.

[bboreham]

Although that may not work work if you actually use the plugin; I've seen Docker get stuck trying to talk to the plugin before starting to accept commands.

[DanielDent]

In a configuration I've been building, I've gone with 4 systemd unit files, plus an environment file, which I've attached.

These files might be helpful for #1836 too.

The only problem with this arrangement is that the weave-plugin service doesn't start cleanly on a reboot because launch-plugin exits uncleanly. Perhaps a "--ifnotlaunched" or similar flag could be made available so that the weave script returns 0 if the plugin is already running.

weave-config.txt
weave-proxy.service.txt
weave-expose.service.txt
weave-plugin.service.txt
weave-router.service.txt

[awh]

Fixed by b3a29be