Open neopointer opened 1 week ago
Why does reset need containerd running anyway?
Because kubelet / containerd don't stop running pods / containers when they get stopped. The kubelet doesn't have any APIs to stop all pods, so k0s reset
relies on containerd to stop containers instead. Therefore, containerd must be up.
you only do reset after stop, so wouldn't it be expected that containerd is offline when you run reset?
In fact, k0s reset
spawns its own containerd instance just for the purpose of shutting down all containers.
I suppose the containerd process that's being spawned fails to start for some reason, hence k0s cannot connect to it.
@twz123 where can I find logs or debug information about this? so that the next time it happens maybe I can collect more info and share here?
Before creating an issue, make sure you've checked the following:
Platform
Version
v1.31.1+k0s.1
Sysinfo
`k0s sysinfo`
What happened?
Reset never works smoothly for me. So I would say I most likely faced different issues so far, but this one is the more commend one
Why does
reset
need containerd running anyway? you only doreset
afterstop
, so wouldn't it be expected thatcontainerd
is offline when you runreset
?This issue could be related to #4211 and #4434.
Steps to reproduce
k0s stop
k0s reset
However, I'm sure this is not reproducible every time. I haven't encountered a way to reproduce the issue reliably.
Expected behavior
No errors during reset.
Actual behavior
Errors during reset, sometimes it hangs for a very long time until it finally exits.
Screenshots and logs
Additional context
No response