John Stoneham created ACCUMULO-1697:
---------------------------------------
Summary: Thread leaks in Tomcat on hot-redeploy
Key: ACCUMULO-1697
URL: https://issues.apache.org/jira/browse/ACCUMULO-1697
Project: Accumulo
Issue Type: Bug
Affects Versions: 1.4.0
Reporter: John Stoneham
Priority: Minor
Testing on 1.3.7 cluster with Tomcat and redeploying my web app by 'touch'ing
the .war file.
In catalina.out:
Sep 06, 2013 4:46:38 PM org.apache.catalina.startup.HostConfig
deleteRedeployResources
INFO: Undeploying context [/myapp]
Sep 06, 2013 4:46:48 PM org.jboss.modcluster.ModClusterService drainSessions
WARN: Failed to drain pending requests from context [/myapp] within specified
timeout: 10 seconds
Sep 06, 2013 4:46:48 PM org.apache.catalina.loader.WebappClassLoader
clearReferencesThreads
SEVERE: The web application [/myapp] appears to have started a thread named
[localhost-startStop-6-SendThread(r03sv04:2181)] but has failed to stop it.
This is very likely to create a memory leak.
Sep 06, 2013 4:46:48 PM org.apache.catalina.loader.WebappClassLoader
clearReferencesThreads
SEVERE: The web application [/myapp] appears to have started a thread named
[localhost-startStop-6-EventThread] but has failed to stop it. This is very
likely to create a memory leak.
Sep 06, 2013 4:46:48 PM org.apache.catalina.loader.WebappClassLoader
clearReferencesThreads
SEVERE: The web application [/myapp] appears to have started a thread named
[Thrift Connection Pool Checker] but has failed to stop it. This is very likely
to create a memory leak.
I can't easily re-test this on a 1.4 or 1.5 cluster at this time, but I
examined the current code in ThriftTransportPool regarding the third message.
No reference is kept to the Daemon instance, so it's never stopped. This is
normally not a problem in JVMs because daemon threads are interrupted
automatically at JVM shutdown, but this is a problem for hot-redeploys in
Tomcat.
It would be nice to have a call to shut down the Thrift checker thread. For the
ZooKeeper threads, we should be able to call ClientCnxn.close() by calling
ZooKeeper.close(), but the ZooKeepers are locked up in ZooSession.sessions and
can't be closed directly.
--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira
