Konnectivity unable to recover after crash

[Wouter0100]

For an unknown reason, our Konnectivity crashes from time to time. We have not found the exact reason for the crash, but after that it Konnectivity is unable to recover and thus keeps on crashing.

The error when it keeps on crashing is as follows:

E1111 11:49:25.377756       1 main.go:49] error: failed to run the frontend server: failed to get uds listener: failed to listen(unix) name /run/konnectivity/konnectivity-server.socket: listen unix /run/konnectivity/konnectivity-server.socket: bind: address already in use

This issue specifically focuses on the fact that Konnectivity is unable to recover after a crash.

[Wouter0100]

This might be a good reason for #598, as that will remove the use of a unix socket. Not sure about other possible workarounds.

[prometherion]

I don't think #598 would solve your issue, mostly because the UDS is required for the API Server and Konnectivity Server interaction.

May I ask you which version of Konnectivity Server are you running on?

[Wouter0100]

For this cluster Konnectivity v0.0.32, due to an older Kubernetes version it seems.

[prometherion]

Until we have enough data to debug, as well as logs, I'm asking if you could try updating konnectivity components version: starting from Kubernetes v1.28.0 there is a release parity with Kubernetes as a main dependency.

You can override the tag by specifying tcp.spec.addons.konnectivity.agent.version and tcp.spec.addons.konnectivity.server.version, using the pattern v0.X.0, where X is the minor Kubernetes version.

You can browse published container images using this link.

[prometherion]

kubernetes-sigs/apiserver-network-proxy#498 seems solving your issue, if you agree I'd close this bug report since a newer version of Konnectivity will fix your issue.