Jaeger: Is there a document to explain the specific meaning of various metrics?

[wuyupengwoaini]

Supporting prometheus metrcis is a good thing for me ,but i find it is difficult to totally understand the real meaning of the metrics exposed by jaeger-agent,jaeger-collector and jaeger-query.

So, I really want to konw whether there is a document to explain the specific meaning of various metrics?

[yurishkuro]

We don't have one at the moment. However, there is a PR that's getting merged soon that adds descriptions to out metrics API, so it would be a good time to add the documentation you are asking for as those descriptions (it will stay in the code and will be up to date). There was also work someone was doing to create standard dashboards.

[wuyupengwoaini]

Could you tell me the PR link?

[objectiser]

The description PR is https://github.com/jaegertracing/jaeger-lib/pull/61.

This PR just adds support for including a description field declaratively in the code - a separate PR would be required to actually define those description values.

[wuyupengwoaini]

Thank you ,waiting for the new PR and the standard dashboards such as grafana dashboards

[jpkrohling]

I know that @Gouthamve is working on a couple of things related to Grafana and we are planning on writing a blog post about how to monitor Jaeger.

Your point about documenting the meaning behind each metric is a great one, but I'm not sure what's the best way to approach this. Have you seen other projects doing this (cc @jstickler)?

[gouthamve]

Well while it's great to do this in the README, it's kinda hard to keep such a list up-to date. Ideally, good HELP text in Prometheus would fix this, and the above PR adds support for that.

See: https://github.com/frodenas/stackdriver_exporter#metrics for inspiration.

[JStickler]

@jpkrohling I'm not sure I have an answer for you, I don't feel like I've got that broad exposure to what other projects are doing their documentation.

But I'm a definite +1 on thinking that documenting the meaning of the various metrics would be helpful.

[yurishkuro]

I suggest we start with the in-code documentation, which makes the metrics self-describing in Prometheus endpoint. Maintaining separate documentation is difficult.