[
https://issues.apache.org/jira/browse/PHOENIX-3728?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16553103#comment-16553103
]
Josh Elser commented on PHOENIX-3728:
-------------------------------------
{code:java}
for (LinkedBlockingQueue<WeakReference<PhoenixConnection>> queue :
connectionQueues) {
queue.clear();
}{code}
You would need to do something like the above instead of
{{connectionQueues.clear()}}, *after* calling close() on
ConnectionQueryServicesImpl (to make sure the executor service stops trying to
run new RenewalThreads).
I am a little concerned about how the redeploy logic in Tomcat works though.
java.sql.DriverManager is going to be caching instances of our PhoenixDriver –
the one we just closed. Are we sure that Tomcat is going to clean up enough
internals to make sure that we get a fresh instance that can actually work?
I would like to see a test along with this change that:
* Connects to phoenix and does some trivial query
* Closes the Phoenix "internals" in the same manner that Tomcat would
* Reconnects to Phoenix in the same manner that Tomcat would
* Verify that some trivial query still works
> renewLeaseExecutor.shutdown() not getting called on PhoenixDriver close()
> method invocation.
> --------------------------------------------------------------------------------------------
>
> Key: PHOENIX-3728
> URL: https://issues.apache.org/jira/browse/PHOENIX-3728
> Project: Phoenix
> Issue Type: Bug
> Reporter: Komal Bansal
> Priority: Blocker
> Attachments: PHOENIX-3728.patch, patch
>
>
> We are using PhoenixDriver for a java service deployed in tomcat. This driver
> creates pool of pre-specified number of threads for lease renewal on first
> connection creation. When the service is redeployed, this thread pool is not
> getting cleared, hence causing memory leaks.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
