BTW, the metric_relabel_config works with below config to avoid 'le="+Inf"' lable, but same configuration with relabel_config does not work.metric_relabel_configs:
- source_labels: [le]
regex: '\+Inf'
action: 'drop'
That is correct. relabel_config is for producing a list of
targets to scrape. metric_relabel_config is used later once the
scape has happened against the metrics returned.
With regards to your request is there a reason for wanting to
remove the metrics with those labels? The example histogram you
gave needs the +Inf le entry to record all work durations above 10
seconds, so removing it will lose information. Also without the
+Inf bucket the histogram_quantile() function will not work (it
would return NaN).
--在2021年2月16日星期二 UTC+8 下午2:35:17<He Wu> 写道:
Hi everyone, I want to collect the kube-apiserver metrics by promethues but drop any label with 'le="+Inf"' , the other labels like 'le="1"' and 'le="10"' need to be kept . I have tried to write relabel_config to achieve , that but did not success. Cloud you please advice how is the relabel_config should be written, thanks!
Here is the sample matric output from kube-apiserver:
'workqueue_work_duration_seconds_bucket{name="non_structural_schema_condition_controller",le="1"} 669workqueue_work_duration_seconds_bucket{name="non_structural_schema_condition_controller",le="10"} 669workqueue_work_duration_seconds_bucket{name="non_structural_schema_condition_controller",le="+Inf"} 669'
Regards,He
You received this message because you are subscribed to the Google Groups "Prometheus Users" group.
To unsubscribe from this group and stop receiving emails from it, send an email to prometheus-use...@googlegroups.com.
To view this discussion on the web visit https://groups.google.com/d/msgid/prometheus-users/7600d164-feb3-4716-b062-7348b7a1e470n%40googlegroups.com.
-- Stuart Clark