We want elasticsearch to autorestart. Previously we configured docker to do this (via docker run auto restart) but this confused openrc because when docker restarted ES the pidfile would change. Figure out some solution that is longer term and makes both things happy.
antarus: on what frequency do the restarts need to be? is there an option that has an init system in the container to restart it vs restarting the entire container?
(In reply to Robin Johnson from comment #1) > antarus: on what frequency do the restarts need to be? is there an option > that has an init system in the container to restart it vs restarting the > entire container? So I see three use cases, perhaps they both are fixed in the current config and we can do nothing (mostly because fixing openrc is very expensive.) 1) Update the config. The config is in the init script, so a "config update" is basically docker stop + docker run, the latter with any updated config options. 2) Update the software. This is a manual operation, and we don't update often so I care less about this use case for packages. 3) Keep it running. Puppet does 3; albeit with a delay. Perhaps the delay is sufficient. Lets let time tell. 2 doesn't matter because we don't do it and we currently expect a human to do it anyway. The current implementation of 1 requires a docker stop / docker run (because docker run is the thing tweaking the config.) This actually works fine today because with restarts off, openrc can track the docker run invocation; so "restart" works. I suspect "reload" doesn't; we may want to check to see what puppet does in this case on a refresh event to call the right one.
We don't run ES on this anymore.