[GitHub] [kafka] robobario commented on pull request #13348: MINOR: Fix ProducerPerformance still counting successful sending when sending failed

["robobario (via GitHub)" <gi...@apache.org>]

robobario commented on PR #13348:
URL: https://github.com/apache/kafka/pull/13348#issuecomment-1471191098

   Hi, I think this introduced a race condition that can cause the producer-perf-test to log a lot of exceptions to console if num records > 500000. Since the `iteration` is incremented from the callback, multiple threads may see the same value when deciding whether to update `latencies`, so it can go out of bounds when updating `latencies`.
   ```
   bin/kafka-producer-perf-test.sh     --topic perf-test     --throughput -1     --num-records 1000000     --producer-props acks=all bootstrap.servers=localhost:9092     --record-size 50
   ```
   Sometimes it logs:
   ```
   [2023-03-16 15:30:57,432] ERROR Error executing user-provided callback on message for topic-partition 'perf-test-0' (org.apache.kafka.clients.producer.internals.ProducerBatch)
   java.lang.ArrayIndexOutOfBoundsException: Index 500001 out of bounds for length 500001
   	at org.apache.kafka.tools.ProducerPerformance$Stats.record(ProducerPerformance.java:38
   ```
   It doesn't kill the test run but it looks alarming to see a couple of hundred errors in console.


-- 
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-unsubscribe@kafka.apache.org

For queries about this service, please contact Infrastructure at:
users@infra.apache.org