[bug] erroneous timeout

[cornfeedhobo]

What happened?:

panic.go:838] "HTTP" verb="GET" URI="/apis/custom.metrics.k8s.io/v1beta1" latency="25.789098ms" userAgent="Go-http-client/2.0" audit-ID="bb032cb7-e6bd-4aee-8d52-3c8ac71d3987" srcIP="10.0.0.151:44714" resp=200
wrap.go:53] timeout or abort while handling: method=GET URI="/apis/custom.metrics.k8s.io/v1beta1" audit-ID="bb032cb7-e6bd-4aee-8d52-3c8ac71d3987"

What did you expect to happen?:

No log spam about a request that reports a successful response (200).

Anything else we need to know?:

• This is still happening in v0.11.2.
• This problem has been going on for nearly 3 years and has gone unaddressed.
• As noted above, and not highlighted in previous issues, it would appear that the request was successful, so why the timeout?
• Previous issues also claimed that the pods default to too small of a resource constraint, which I've also ruled out by removing all resource constraints and placing on an over powered node. As well, this doesn't make sense considering there are popular charts like rancher-monitoring that add many PrometheusRule resources.

Known workarounds?:

It is possible to set rules.default=false, which disables the baseline metrics instead of addressing the underlying issue.

Environment:

• prometheus-adapter version: v0.11.2
• prometheus version: v2.38.0
• Kubernetes version (use kubectl version): v1.26.10+k3s2

[dgrisonnet]

/triage accepted /assign

[cornfeedhobo]

I'll post some debug logs when I create another test cluster.

[dgrisonnet]

Which verbose level are you using?

[cornfeedhobo]

I'll do my best to revisit this, but unfortunately I had to move on to other things and don't know when I'll have capacity to bring up a test cluster again. I'm frankly running out of resources.