should probes behave more like kubernetes probes?

[adrian-gierakowski]

looks like currently failing readiness_probe in PC would cause process to be killed (and restarted, depending on availability config), however I believe only failing liveness probe should lead to this outcome. readiness_probe should only be used when managing startup of other process which require given process to be ready first

[F1bonacc1]

Hey @adrian-gierakowski,
I more than agree with you.
I absent-mindedly swapped between them when I first added the health probes, and wanted to switch them ever since.
The only reason I didn't do it is to keep backward compatibility with PC.
If everyone agrees that K8s compatibility wins over backward compatibility, I will gladly switch between the 2:

1. Readiness will be used only to indicate initial readiness for other processes (PC won't be able to withhold the incoming traffic like in K8s).
2. Liveness will restart if the probe fails.

[adrian-gierakowski]

Maybe we should put it behind a flag for now and then swap the default behaviour in the next major? I believe eventually a major bump would be desirable, since there are other defaults that would be good to change, like disable_env_expansion (which should be off by default due to it's un-intuitive behaviour).