[jira] [Commented] (KAFKA-4386) Producer Metrics Explanation

["huxi (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/KAFKA-4386?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15646087#comment-15646087 ] 

huxi commented on KAFKA-4386:
-----------------------------

In my understanding, these metrics are all request-level. Take request latency as example, it measure the average time of request latency, which is the difference between the response received time and  request creation time. As for the multiple producers, each producer has its own client id, and in turn has its own measure.

> Producer Metrics Explanation
> ----------------------------
>
>                 Key: KAFKA-4386
>                 URL: https://issues.apache.org/jira/browse/KAFKA-4386
>             Project: Kafka
>          Issue Type: Bug
>          Components: producer 
>            Reporter: Pratik kumar
>              Labels: producer
>
> Context :
> Kafka Producer 0.8.x
> Problem:
> Kafka Producer emits metrics regarding request size stats, request latency and request rate stats.
> But the inherent meaning of the these metrics are not clear. What does this measure?
> Is for each producer send request(which contains batches of messages per broker)? OR Is it for a batch of messages defined according to user batching policy? What happens when some application code has multiple async producers to increase performance (how are rate and percentiles measured?)?



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)