This message was deleted.

Note that not knowing if you're using k3d or RKE1 or what makes this a bit of a shot in the dark (if you're using k3s or RKE2 directly, then docker daemon doesn't matter). I'm not 100% certain, but my gut reaction is to tell you to restart your kube-apiserver . By default, Docker will keep any containers it has running while the service restarts (there's an option you can set in the /etc/docker/daemon.json to change the behavior, I think it's live-reload but Docker docs will tell you for certain). While the container keeps running, open file handles or network sockets may end up stale and a lot of containerized software just assumes a restart rather than coding in recovery. I think one time I saw the container lose the network interface altogether. So that may be what's happening to you.

Hello @rough-farmer-49135 thank you for you hints. I'm using RKE1. I have already tried to restart the docker service a few times, which restarts also the rke1 kube-api container, but no luck. Not sure what else I can do. Happy to hear some more suggestions. thank you in advance!

I'd explicitly set live-restore (just looked it up - https://docs.docker.com/config/containers/live-restore/ ) so it does what you expect just in general. For RKE1, while I've never used it, if you're setting live-restore to false, I think I'd use the RKE1 scripts/binaries to shut down the Kubernetes cluster, restart docker daemon, then use RKE1 scripts/binaries to bring it back up. Distributed decision-making takes a while to settle, and sometimes if things are started up out of expected order you can get into deadlocked loops and never have it start.

Ah now I see I will try to check my

/etc/docker/daemon.json

! Thank you will report back!

A year ago I was using RKE2 at a different job (security was being emphasized, so seemed a definitely better choice). Current job is using k3d (which runs k3s in a docker container, and it's just a handy development/demonstration environment for eventual resource constrained systems, so seems like a good choice too).

Thank you for the Information