worker node shutdown waits long for containerd-shim

[matti]

waits for containerd-shims:

(Ubuntu 20.04, running in proxmox vm, hence the screenshot)

[jnummelin]

worker node shutdown waits long

How long is long? :) I'd guess it's around 90sec which I remember is the systemd default

[matti]

but from users point of view: I have nothing running, so in my opinion a node should reboot/shut down fast

[jnummelin]

The root cause is containerd shims not handling SIGTERM properly: containerd/containerd#5502

A fix upstream is targeted for containerd 1.6, there's not much we can do on k0s level before that ships out.

[github-actions]

The issue is marked as stale since no activity has been recorded in 30 days

[jnummelin]

Remove stale, we're still waiting for upstream to have fix for this

[twz123]

The containerd fix has been backported into 1.5.x recently (containerd/containerd#6509), so maybe there'll be a new release soon.

[jnummelin]

There's now 1.6.0 containerd release which has this fix. We should bump main to use that.

No release yet for 1.5.x that includes this fix.

[github-actions]

The issue is marked as stale since no activity has been recorded in 30 days

[matti]

asdf

[twz123]

k0s v1.23.5+k0s.0 is out, shipping containerd 1.5.11, which has the fix.

[github-actions]

The issue is marked as stale since no activity has been recorded in 30 days

[twz123]

Closing as I believe this is fixed as of v1.23.5+k0s.0. Please ping me if it's not working as intended.