[
https://issues.apache.org/jira/browse/KAFKA-7734?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16721658#comment-16721658
]
John Roesler commented on KAFKA-7734:
-------------------------------------
[~guozhang],
I've already made this comment on the PR, but it seems more manageable to
establish a contract that says that the MetricName will always contain the tags
in alphanumeric order. This could be implemented more easily, simply by storing
the tags in a TreeMap in MetricName, and wouldn't require any other changes to
the code, or incur future maintenence costs.
WDYT?
> Metrics tags should use LinkedHashMap to guarantee ordering
> -----------------------------------------------------------
>
> Key: KAFKA-7734
> URL: https://issues.apache.org/jira/browse/KAFKA-7734
> Project: Kafka
> Issue Type: Improvement
> Components: metrics
> Reporter: Guozhang Wang
> Assignee: lambdaliu
> Priority: Major
>
> Today we store metrics tags inside MetricName from various places, and many
> of them are using `HashMap`. However, for metrics reporters like JMXReporter,
> the mBeanName is constructed by looping over `metricName.tags().entrySet()`
> which does not guarantee ordering. This resulted a few places where the
> mBeanName string not as expected, e.g. we document the Streams cache metrics
> as
> {code}
> kafka.streams:type=stream-record-cache-metrics,client-id=([-.\w]+),task-id=([-.\w]+),record-cache-id=([-.\w]+)
> {code}
> However, what I've seen from JMXReporter is, for example:
> {code}
> kafka.streams:type=stream-record-cache-metrics,record-cache-id=all,client-id=streams-saak-test-client-StreamThread-1,task-id=1_3
> {code}
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
