Improving the systemd unit file example

[l-x]

We recently came across the problem that systemd stop roadrunner.service with the unit file shown at https://docs.roadrunner.dev/docs/app-server/systemd sends both the Roadrunner process and the PHP worker processes SIGTERM at the same time. In our situation, this led to Roadrunner immediately restarting all workers and then waiting for them to finish - which of course never happened. After 30 seconds, Roadrunner then forcefully terminated these new workers.

We were able to solve this by setting the value for KillMode to mixed in the unit file. In our case, we also had to adjust the value for TimeoutStopSec to the Roadrunner configuration for jobs.pool.destroy_timeout.

Perhaps it makes sense to include these findings in the example unit file.

[rustatian]

Hey @l-x 👋🏻
Yeah, great finding. Would you mind sending a PR to improve docs? (roadrunner-server/docs repo)

[l-x]

If a service of Type=notify/Type=notify-reload sends "EXTEND_TIMEOUT_USEC=…", this may cause the stop time to be extended beyond TimeoutStopSec=. The first receipt of this message must occur before TimeoutStopSec= is exceeded, and once the stop time has extended beyond TimeoutStopSec=, the service manager will allow the service to continue to stop, provided the service repeats "EXTEND_TIMEOUT_USEC=…" within the interval specified, or terminates itself (see sd_notify(3)).

How does Roadrunner behave in this situations? Maybe actually adjusting TimeoutStopSec is not required...

[rustatian]

RR does not handle these systemd events.