Every so often I get:
time=2026-09-21T20:06:03.013Z level=ERROR source=http.go:231 msg="error encoding and sending metric family: write tcp [::1]:9100->[::1]:56472: write: broken pipe
Once it gets into this state, it remains unresponsive to metric requests.
While I've not traced down the overall source, restarting the pod makes the problem go away.
Can there be a flag added to node-exporter to terminate the pod with an error if that error message pops up? Kubernetes will reschedule it and I can try and sort my environment based on any events/patterns that pop up.
Every so often I get:
Once it gets into this state, it remains unresponsive to metric requests.
While I've not traced down the overall source, restarting the pod makes the problem go away.
Can there be a flag added to node-exporter to terminate the pod with an error if that error message pops up? Kubernetes will reschedule it and I can try and sort my environment based on any events/patterns that pop up.