megfigura commented on issue #7304: URL: https://github.com/apache/pulsar/issues/7304#issuecomment-854943245
Also hit this issue with a similar use-case (creating batches for a Spark application). The issue is still present in the latest 2.7 branch as of commit 6bbb8d5c34d59a3a2cdd88e3e091de37b60d274a, though the leak I see is much slower than the one pictured in the screenshot above, so I wonder if the problem has been partially fixed since 2.5.2. It's easy to reproduce, even with curl. e.g. `while true; do curl http://<admin IP:port>/admin/v2/persistent/public/default/some_topic/lastMessageId; done` If you enable Netty leak detection (`-Dpulsar.allocator.leak_detection=Advanced -Dio.netty.leakDetectionLevel=advanced -Dio.netty.leakDetection.targetRecords=30`) LEAK messages start showing up almost immediately. After the first few messages, they seem to show up more consistently if data is being produced to the topic, though it will still happen regardless. Maybe producing data just causes GCs to happen more consistently so the Netty leak detection kicks in. -- 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. For queries about this service, please contact Infrastructure at: us...@infra.apache.org
