Metrics: possibly missing labels in internal_metrics documentation

[zamazan4ik]

A note for the community

• Please vote on this issue by adding a 👍 reaction to the original issue to help the community and maintainers prioritize this request
• If you are interested in working on this issue or have submitted a pull request, please leave a comment

Problem

According to the documentation, both collect_completed_total, collect_duration_seconds, events_discarded_total, http_error_response_total, http_request_errors_total, parse_errors_total, request_errors_total have the following description: "... for this component.". But in the labels for these metrics there are only host and pid. According to the description, labels like component_id, component_kind are missing.

Documentation link: https://vector.dev/docs/reference/configuration/sources/internal_metrics

Configuration

No response

Version

Latest master branch

Debug Output

No response

Example Data

No response

Additional Context

No response

References

No response

[zamazan4ik]

Another one: decode_errors_total and encode_errors_total have different set of labels. Is it true?

[zamazan4ik]

For internal_metrics_cardinality and internal_metrics_cardinality_total at least host label is missing in the documentation (pid could be also a case here). With the current Vector version in Prometheus metrics I have the following:

# HELP vector_internal_metrics_cardinality internal_metrics_cardinality
# TYPE vector_internal_metrics_cardinality gauge
vector_internal_metrics_cardinality{host="MacBook-Pro-Alexander.local"} 73 1663550075574
# HELP vector_internal_metrics_cardinality_total internal_metrics_cardinality_total
# TYPE vector_internal_metrics_cardinality_total counter
vector_internal_metrics_cardinality_total{host="MacBook-Pro-Alexander.local"} 73 1663550075574

[zamazan4ik]

Not sure about request_duration_seconds - please recheck it :) I have a feeling (possibly wrong) what it also should have more labels.

[zamazan4ik]

@bruceg probably you could help here :)

[jszwedko]

According to the documentation, both collect_completed_total, collect_duration_seconds, events_discarded_total, http_error_response_total, http_request_errors_total, parse_errors_total, request_errors_total have the following description: "... for this component.". But in the labels for these metrics there are only host and pid. According to the description, labels like component_id, component_kind are missing.

Hmm, yes the metrics you identified should have component labels. Are you not seeing those for a specific component?

vector_internal_metrics_cardinality and vector_internal_metrics_cardinality_total should only have the host tag by default. pid is optional and needs to be opted into (https://vector.dev/docs/reference/configuration/sources/internal_metrics/#tags.pid_key).

[zamazan4ik]

Hmm, yes the metrics you identified should have component labels. Are you not seeing those for a specific component?

I do not see component-related labels in the documentation. E.g. this one: https://vector.dev/docs/reference/configuration/sources/internal_metrics/#collect_completed_total

vector_internal_metrics_cardinality and vector_internal_metrics_cardinality_total should only have the host tag by default.

Yeah, but the documentation doesn't show even host tag: https://vector.dev/docs/reference/configuration/sources/internal_metrics/#internal_metrics_cardinality

My whole issue is just about incomplete/confusing documentation, not aboout an actual metrics behaviour. I found these documentation mistakes during the work on the Grafana dashboard for Vector.

[zamazan4ik]

Also, I have found some inconsistencies in metrics names. For Kafka almost all metrics have kafka_ prefixes. But there are metrics like consumer_offset_updates_failed_total and events_failed_total which correspond to Kafka but have no kafka_ prefix. I think it should be fixed, hopefully in some backward-compatible way.

[zamazan4ik]

@jszwedko I think I cannot provide a PR with fixing the issue with wrong labels for the metrics in the documentation since I don't really know, for which exactly metrics I should fix it. So probably it should be assigned to someone from Vector dev team.

[zamazan4ik]

@jszwedko @bruceg maybe you have any update?

[zamazan4ik]

@jszwedko friendly reminder about the issue :)

[jszwedko]

Hey! Sorry, this is in our backlog. We just haven't gotten to it yet. Thanks for filing it though!