Add Restart=on-failure in systemd units firebird and firebird-superserver

[basid-irk]

Linked with document how OOM killed processes interact with Restart=on-abort/on-abnormal.
After OOMKill firebird process on Debian(?), systemd may leave unit's firebird or firebird-superserver in exited/failed state and fbguard autorestart feature will not work.
As documented in systemd.service Restart=on-failure make autorestart for unit with OOMKilled process.

[AlexPeshkoff]

Very strange that guardian does not work - I doubt this tiny process may get killed with OOM killer. May be SE-Linux jokes here too? Do you have a reproducible test case?

[basid-irk]

No. Only (private) study of the problem after the fact.
My conclusion based on combination of symptoms and after search for "how to restart systemd unit after OOMKill".

[basid-irk]

I doubt this tiny process may get killed with OOM killer.

OOMKill-ed process firebird, but (highly likely) systemd stopped unit as whole and leave it as exited/failed.
No more precise conclusions, since the client rebooted the host before my study.

[tomaszdubiel18]

I can confirm this. It happens among our customers on Debian and Ubuntu. OOM killer and then service is stopped. You need to manually start it.

[AlexPeshkoff]

Please check it in fb6