Closed marevers closed 3 months ago
Thanks, nice catch. However, this might be on purpouse due conversion from OTEL to Prometheus metrics. is that right @mariomac ?
There were more (http_client and rpc_client) metrics with incorrect names. I fixed those too.
Thank you for your contribution @marevers !
Actually, the dashboard was composed for an scenario where, between Beyla and Grafana, there was the Grafana Agent doing the OpenTelemetry ingestion and transforming the metrics to Mimir by using a given naming convention.
It seems that the main scenario is that our users send the data directly to the OpenTelemetry endpoint, which uses a different naming convention to transform the metrics, so we will internally discuss wether we should adopt the metrics that you are proposing and come back to this PR with a decision.
All modified and coverable lines are covered by tests :white_check_mark:
Project coverage is 39.25%. Comparing base (
490a82c
) to head (70dbacb
). Report is 1 commits behind head on main.
:umbrella: View full report in Codecov by Sentry.
:loudspeaker: Have feedback on the report? Share it here.
It seems that the main scenario is that our users send the data directly to the OpenTelemetry endpoint, which uses a different naming convention to transform the metrics, so we will internally discuss wether we should adopt the metrics that you are proposing and come back to this PR with a decision.
@mariomac Thank you for responding. Upon a deeper look I also realized some label names are incorrect which is causing some sum by
functions to not work. Should you accept this PR then I can fix that as well.
After discussing internally, we will default the names as you propose in this PR. Thanks for your contribution @marevers !
@mariomac thank you. I will open a separate PR for the corrections in the label names.
This PR fixes incorrect metric names in the Grafana dashboard.
Metrics names did not include
_seconds_
e.g. http_server_request_duration_bucket should actually be http_server_request_duration_seconds_bucket and rpc_server_duration_bucket should be rpc_server_duration_seconds_bucket.With these changes, the dashboard works out of the box. I assume the metric names were changed at some point, but the dashboard was not changed along with it accordingly.