Stopping/restarting doesn't actually stop/restart postgresql

[glinton]

root@data.db.3.1:/ # sv stop db
ok: down: db: 0s, normally up, want up
root@data.db.3.1:/ # ps aux | grep postgre
gonano      61  0.0  0.6 2725420 25416 ?       Ss   17:09   0:00 postgres: database 192.168.0.20(41770) idle

[notxarb]

I think postgresql doesn't like signals, it would rather have pgctl start/stop/restart it.

[glinton]

Could this be what we need - https://serverfault.com/a/410339

[tylerflint]

@glinton wow, great find!

[notxarb]

Another thing that I found is that we want to use chpst rather than su to change user. It looks like a lot of the services are using su right now. Using su creates an intermediary process between runit and the actual service process. When runit goes to stop the service, it kills the su process which then abandons the service process rather than shutting it down.

[glinton]

It looks like there are still issues when using chpst. Apparently, one app reports issues with postgres every deploy.