You are viewing a plain text version of this content. The canonical link for it is here.
Posted to jira@kafka.apache.org by "Ricardo Bartolome (JIRA)" <ji...@apache.org> on 2017/12/13 09:51:00 UTC

[jira] [Created] (KAFKA-6358) Per topic producer/fetch_consumer/fetch_follower metrics

Ricardo Bartolome created KAFKA-6358:
----------------------------------------

             Summary: Per topic producer/fetch_consumer/fetch_follower metrics
                 Key: KAFKA-6358
                 URL: https://issues.apache.org/jira/browse/KAFKA-6358
             Project: Kafka
          Issue Type: Wish
          Components: metrics
    Affects Versions: 1.0.0
            Reporter: Ricardo Bartolome
            Priority: Minor


We are using the following JMX beans to monitor Kafka 1.0.0:

{code}
kafka.network:type=RequestMetrics,name=TotalTimeMs,request=Produce
    Mean
    50thPercentile
    ...
    99thPercentile

kafka.network:type=RequestMetrics,name=RequestsPerSec,request=FetchConsumer
    Count

kafka.network:type=RequestMetrics,name=RequestsPerSec,request=FetchFollower
    Count
{code}

There are more, but this provide an idea of what we are using in order to get produce/fetch operations on a per-broker basis. Nevertheless, in order to identify abusing consumers/clients in our kafka cluster, we would appreciate to have these metrics in a per-topic basis.

As example of per-topic metrics we have:
{code}
kafka.server:type=BrokerTopicMetrics,name=BytesOutPerSec,topic=.*
{code}

Where we have a per-topic bean with a "Count" attribute that we can query. That way we can know which topics are ingesting more data and which ones less data. We can't do that with the metric explained above.

Would you consider a change in an upcoming Kafka version as a feature request?



--
This message was sent by Atlassian JIRA
(v6.4.14#64029)