feature: systemd overrides for rke2-server and agents

[michalg91]

Summary

We've got some large master nodes which are having large etcd database and during upgrade when rolling restart occurs there can be timeouts on restarting service. It will fail on node and start restarting next one, which is not wanted.

what we can do here is:

• ignore this error and add another task with retry to check if unit is started
• add ability to provide override.conf for service with eg. extended timeout (TimeoutSec,TimeoutStartSec).

Issue Type

Feature Idea

[MonolithProjects]

Hi @michalg91, according my knowledge rke2-server service has TimeoutStartSec=0 so there is no timeout. But I think that failed rke2-server restart should actually stop the whole playbook and not to continue to another node - so this is something which could be improved.

[intonet]

I have this problem too — almost one-third of my clusters fail when restarting the rke2-server on one of their masters. (The services themselves are fine; the restart just takes longer.) Maybe we need to add TimeoutStopSec, or, as @michalg91 suggests, ignore the error here and wait for the service to start again. This failing task without waiting or stop playbook causes some times race condition on helm operator when one master still is restarting a second one is restarted too.