You are viewing a plain text version of this content. The canonical link for it is here.

Posted to issues@zookeeper.apache.org by "xiongjianbo (Jira)" <ji...@apache.org> on 2022/03/28 08:35:00 UTC

[jira] [Created] (ZOOKEEPER-4502) SyncRequestProcessor leak when leader election occurred

xiongjianbo created ZOOKEEPER-4502:
--------------------------------------

             Summary: SyncRequestProcessor leak when leader election occurred
                 Key: ZOOKEEPER-4502
                 URL: https://issues.apache.org/jira/browse/ZOOKEEPER-4502
             Project: ZooKeeper
          Issue Type: Bug
          Components: quorum
    Affects Versions: 3.6.3
            Reporter: xiongjianbo


After a three-node ZooKeeper cluster runs for a period of time, OutOfMemory occurs. Check the memory stack and find that the number of SyncRequestProcessor objects is over 400.

!image-2022-03-28-16-18-43-147.png!

 

Check the previous modification records. It is found that the Learner.shutdown logic is modified this time.

[https://github.com/apache/zookeeper/pull/1619]

After the modification, the original invoking logic is changed.

Before modification: Learner.shutdown() -> LearnerZooKeeperServer.shutdown() -> ZooKeeperServer.shutdown()

After modification: Learner.shutdown() -> ZooKeeperServer.shutdown(boolean)

Finally, LearnerZooKeeperServer.syncProcessor.shutdown() was never called.

syncProcessor



--
This message was sent by Atlassian Jira
(v8.20.1#820001)