search

signalfx / signalfx-k8s-metrics-adapter

Horizontal Pod Autoscaler custom/external metrics provider for Kubernetes that uses SignalFx as the backend
Apache License 2.0
10 stars 15 forks source link

Loglevel not completely respected? #13

Closed rjmsilveira closed 3 years ago

rjmsilveira commented 4 years ago

Hi, I followed Klog log levels, and was trying to reduce output of adapter. I set the value to 1, and later to 0, and I am still seeing this (below) requested constantly (multiple times per second) I1023 14:17:03.802812 1 provider.go:169] ListAllMetrics() -> []

Is this the normal expected behavior, and if so, how can we stop seeing that verbosity?

Thanks,

keitwb commented 4 years ago

I think the Helm chart would need to be enhanced to add a cli flag minloglevel and allow that to be configurable (the default is the lowest level). If you make a PR for that I would gladly accept it.

keitwb commented 4 years ago

The logs you are seeing come from glog and are from the k8s metrics adapter library that this app uses internally. The possible flags are documented in the link.

rjmsilveira commented 4 years ago

Thanks for the reply.

I also took a look where this string was coming from and I noticed that https://github.com/signalfx/signalfx-k8s-metrics-adapter/blob/52d5080d63829bb1c43745a7506b01cda4c95be7/internal/provider.go#L169 shows it has no level like V(5). klog.V(5).Infof("ListAllMetrics() -> %v", out) I believe this is the reason why it is showing this log no matter the value introduced.