How to set log level for metric server

[karimrut]

Hi!

How do I set the log level for the Keda metric server to warn or error? Not sure what it means to set "an integer value greater than 0". Any value between 1-3 still gives me info level logging.

Ref: https://github.com/kedacore/charts/tree/main/keda
logging.metricServer.level

logging:
  operator:
    level: error
  metricServer:
    level: ? #0=info, 4=debug

Thanks in advance
Karim

[JorTurFer]

You can check values in klog repo: https://github.com/kubernetes/klog/blob/main/internal/severity/severity.go#L30

[tomkerkhove]

Updating Chart: kedacore/charts#323

[wiseelf]

Okay, checked the values in repo, 1-3 still gives INFO level. And you can't provide there string value like "WARNING"

[JorTurFer]

Hi, no you can't provide the value using string, you have to use numbers, from 0 to 4.

1-3 still gives INFO level

Could you share your KEDA version and the logs that shouldn't be there to check them?

[kbocock-krg]

I am baffled that something so simple (logging level) has been made overly complex and deviates from consistent standards.

[pandvag]

Hi everyone,

we are still facing the same problem ... all logs are fullfilled with info logging like

"metrics"={"metadata":{},"items":[{"metricName":"s0-http-XYZ_002Fkeda-add-ons-http-interceptor","metricLabels":null,"timestamp":"2024-11-12T09:27:19Z","value":"0"}]} "scaledObjectName"="keda-add-ons-http-interceptor" "scaledObjectNamespace"="XYZ"

How is it possible to have only error messages printed ?

keda v2.16.0 on aks v1.30.5

Thanks for any hints
Matthias

[pandvag]

I could get rid of info logging by using loglevel 0 (haven't used that setting, because it was explizit stated as info logging) ... is it supposed to be like that ? If so, please clarify the docs in a spare minute ...

[JorTurFer]

you can configure the values in helm chart https://github.com/kedacore/charts/blob/3ad2830549dd9faf52da37dc833d5102ac3fb36b/keda/values.yaml#L373-L402

[pandvag]

i know @JorTurFer, but error level is 0 ... all other log-levels are at least printing info ... doc is stating log-level 0 is info .. that seams not correct

[JorTurFer]

There are 2 issues related with incorrect level logs in the metrics server:

• automaxprocs message in the beginning -> fix: ensure consistent JSON log format for automaxprocs #6335
• Some upstream tracing logs in the middle of the executions

Is any of them?

[pandvag]

i only can confirm that by using loglevel 0 within keda-operator-metrics-apiserver we are only having error logs printed