hudeqi commented on PR #14553: URL: https://github.com/apache/kafka/pull/14553#issuecomment-1778612496
> > I tested this process on the latest trunk branch and it did not reproduce. My process is: create a topic with 1 partition, then write traffic and stop writing. For example, the current isr is [2,0,1], I will submit a reassign, and the isr is [2,0,1] , 2 of which logdir changes. Finally execute reassign. Please confirm whether my testing process is different from yours? Thanks @drawxy > > Hi @hudeqi, in your test, the replica 2 already was the leader and it aleady kept the fetch state of the other followers. And the ISR set wouldn't shrink without produce traffic due to there was no lag from leader and follower (the leader replica 2 was aware of the log end offset of other followers). For example, 1. you can create a topic with 1 partition and the assignment list is [0,1] (make replica 0 to be the leader); 2. produce some messages to the partition and stop; 3. submit a partition reassignment [2,1]; 4. submit a log dir alter for replica 2. After that, you can see the log file is fullly filled. > > Btw, I close this PR by mistakes, could you help to reactive it, thanks! Can you pull the latest trunk branch and reproduce it? The relevant logs you mentioned have been modified on the latest trunk. And I tested it multiple times based on the steps you gave based on the trunk branch, and everything seems to be normal. -- 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: jira-unsubscr...@kafka.apache.org For queries about this service, please contact Infrastructure at: us...@infra.apache.org
