KAFKA-19865: Document queues metrics changes in ops.html

[JimmyWang6]

From
KAFKA-19865，
This PR introduces documentation support for monitoring Kafka share
groups and share consumers, including new metric registries and
documentation updates. The changes add new classes to define and expose
share group metrics, and update the documentation to describe these
metrics and how to monitor them.

Pre-view:

[JimmyWang6]

Not quite sure if we need to show the DelayedOperationPurgatory metrics in the documentation, as it doesn't only apply to share group.

[apoorvmittal10]

I think we should mention this. I do only see that following is defined in Kafka documentation kafka.server:type=DelayedRemoteFetchMetrics,name=ExpiresPerSec,topic=([-.\w]+) and not for others. Also I see kafka.server:type=DelayedOperationPurgatory,name=PurgatorySize,delayedOperation has been declared separately for Produce and Fetch then we should follow the similar pattern.

[JimmyWang6]

Got it. It's updated.

[JimmyWang6]

Hi @apoorvmittal10 @AndrewJSchofield, gentle ping on this. Please take a look when you get a chance. Thanks!

[apoorvmittal10]

Thanks for the PR @JimmyWang6. mbean miss the metric name. In Kafka documentation I can see other metrics do define the mbean correctly.

[apoorvmittal10]

In the PR we have just client metrics as sub resource under Share Group Montioring, but should we have coordinator as well which can contain group coordinator and share coordinator metrics. The other reason for this suggestion is that coordinator metrics are not yammer and broker ones are, hence the metrics name casing differs. Hence the table of metrics look odd with differnt casing so if we have the respective sections then it ll make sense. @AndrewJSchofield wdyt?

[AndrewJSchofield]

In the PR we have just client metrics as sub resource under Share Group Montioring, but should we have coordinator as well which can contain group coordinator and share coordinator metrics. The other reason for this suggestion is that coordinator metrics are not yammer and broker ones are, hence the metrics name casing differs. Hence the table of metrics look odd with differnt casing so if we have the respective sections then it ll make sense. @AndrewJSchofield wdyt?

I agree. It looks a bit odd with the different types of metrics mixed up together.

[apoorvmittal10]

@JimmyWang6 Can we please update the description to how documentation looks now after the changes?

[JimmyWang6]

Updated the document and description based on the feedback, thanks:)