This page is meant as a template for writing a KIP. To create a KIP choose Tools->Copy on this page and modify with your content and replace the heading with the next KIP number and a description of your issue. Replace anything in italics with your own description.
Current state: [One of "Under Discussion", "Accepted", "Rejected"]
Discussion thread: here [Change the link from the KIP proposal email archive to your own email thread]
JIRA:
Please keep the discussion on the mailing list rather than commenting on the wiki (wiki discussions get unwieldy fast).
The kafka-consumer-perf-test tool is a great tool for quickly assessing raw consumer performance. Currently, it subscribes to all partitions and provides an overview of the entire cluster's performance. However, the tool does not support testing the performance of a single broker or partition. Introducing this capability would be highly beneficial for debugging and estimating capacity and resource requirements.
In a real-world scenario, our production environment encountered an issue where a Spark Structured Streaming consumer was experiencing lag. This lag was due to some tasks taking longer to complete. Despite all tasks reading from different partitions with an equal number of records to process, the fetch-time for some partitions was noticeably higher than for others. By modifying the kafka-consumer-perf-test tool to test individual partitions, we were able to isolate the problem to a single broker with a faulty switch.
This modification provided the Hardware team with a straightforward method to conduct experiments and verify resolutions, ultimately streamlining the troubleshooting process and ensuring smoother operations.
Add following new optional parameters to kafka-consumer-perf-test:
Describe the new thing you want to do in appropriate detail. This may be fairly extensive and have large subsections of its own. Or it may be a few sentences. Use judgement based on the scope of the change.
Describe in few sentences how the KIP will be tested. We are mostly interested in system tests (since unit-tests are specific to implementation details). How will we know that the implementation works as expected? How will we know nothing broke?
If there are alternative ways of accomplishing the same thing, what were they? The purpose of this section is to motivate why the design is the way it is and not some other way.