Abnormal Linux data collected by MAL

[brave-programmer]

otel-rules/vm.yaml config:

• name: cpu_total_percentage
  exp: (node_cpu_seconds_total * 100).tagNotEqual('mode' , 'idle').sum(['node_identifier_host_name']).rate('PT1M')
• name: cpu_total_percentage_idle
  exp: (node_cpu_seconds_total * 100).tagEqual('mode' , 'idle').sum(['node_identifier_host_name']).rate('PT1M')

dashboard show:
cpu_total_percentage percentage is ok.
cpu_total_percentage_idle percentage is too big???

[wu-sheng]

Could you write in a complete sentence? cpu_total_percentage_idle percentage is too big??? this doesn't make sense. Hard to understand.

[brave-programmer]

A two core Linux, theoretically, the sum of idle and non idle should be equal to 200%, but it shows 30% non idle and 2328954% idle

[wu-sheng]

I don't think we have cpu_total_percentage_idle. If you are writing the MAL, you should check the RAW data reported by OTEL Linux exporter.
We don't have your data or env, it is hard to say.

[brave-programmer]

I have tried multiple Linux platforms and the results are all abnormal. But it is normal to use PromQL statistics in Prometheus.
you can try it. add config on otel-rules/vm.yaml. and show on dashboard(Can monitor any LINUX).

• name: cpu_total_percentage_idle
  exp: (node_cpu_seconds_total * 100).tagEqual('mode' , 'idle').sum(['node_identifier_host_name']).rate('PT1M')

[wu-sheng]

I would not try. Sorry. We have thousands of metrics reported and processed through various features.
I am unaffordable to test one by one, especially to test something we don't declare about.

[brave-programmer]

ok, thanks!