kpoof daemon just stop working sometimes

[gerbil]

I can see random connection issues with kpood daemon on my local kuber cluster, without any reasons.. it just stops working so i need to restart it.
any ideas how can i troubleshoot this?

[lypht]

Are other long-running processes also timing out? Is the local cluster running on a laptop with any power-saving mode (clam-shell close, for example)? The daemon doesn't withstand power cycles or similar OS-governed processes.

[gerbil]

Nope, it's centos based cluster on vms in my dc; All kubeletes running with connection timeout=0

[gerbil]

Weird stuff, sometime in the middle of the night:
E0503 02:16:58.605281 5471 portforward.go:233] lost connection to pod

How can i detect those failures and fix them automagically?

[lypht]

By any chance are you doing large pipes to the pod, such as from a mongodump or similar? If the throughput is greater than the pod can handle, it will close the connection.