Add plugin for k3s server and agent

[slaecker]

For K3s servers the plugin checks k3s.service and cluster availability using kubectl. For K3s agents k3s-agent.service is being checked to be active.

[thkukuk]

Do we have meanwhile k3s RPMs which get installed into /usr?
If k3s is installed outside of /usr, so /usr/local or /opt, a rollback would only revert the OS changes, but not k3s itself. So what happend would be:

1. k3s gets updated
2. k3s does not come up correct at next reboot, health-checker does rollback
3. Old OS (includes the old bugs and security problems) boots, updated k3s is started
4. updated k3s does still not start (rollback not possible, since k3s is not part of the transaction)
5. health-checker will stop machine

Is this really what we want?

[slaecker]

Valid points. My usecase was that a recent kernel upgrade broke K3s. In this case a rollback fixed the issue.

I haven't taken into account k3s being installed via package. I always deploy using Ansible. In this case the cluster "should" be in a running state before a transactional update.

Looks like the plugin is covering an edge case not suited for general use.