chickenchickenlove opened a new pull request, #24134:
URL: https://github.com/apache/kafka/pull/24134

   When a Streams client constructs a task’s input partitions using local 
metadata, that metadata may differ from what the broker used to compute the 
assignment.
   
   For example, metadata fetched from a broker whose metadata view is lagging 
may contain a different partition count. Refreshing client metadata alone 
cannot establish which partition count was used when the assignment was 
computed.
   
   This PR is a PoC that stores a topic metadata snapshot associated with the 
assignment on the broker and includes it in heartbeat responses. The client 
uses the broker-provided partition counts to construct active tasks’ input 
partitions. Task ID remains the unit of assignment and ownership reporting.
   
   This draft is intended to support community discussion and the development 
of a KIP. The response currently includes topic names, topic IDs, and partition 
counts. The client processes this information through metadata resolution, 
subscription state updates, and callback execution.
   
   The implementation is still preliminary. Concurrency, snapshot consistency, 
compatibility, and edge cases require further review. Tests have not yet been 
updated, and end-to-end validation has not been performed.


-- 
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.

To unsubscribe, e-mail: [email protected]

For queries about this service, please contact Infrastructure at:
[email protected]

  • [PR] KAFKA-21188: Prototype brok... chickenchickenlove (via GitHub) via jira

Reply via email to