[
https://issues.apache.org/jira/browse/PHOENIX-1403?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Gabriel Reid updated PHOENIX-1403:
----------------------------------
Attachment: PHOENIX-1403.patch
Trivial patch to resolve the issue. I haven't added an integration test to test
this, as the failing test condition basically causes a denial of service on ZK
(leaving a bunch of leaked connections around), but I have tested this patch to
ensure it does what it's supposed to.
> Race condition can cause ConnectionQueryServicesImpl to be re-initialized
> -------------------------------------------------------------------------
>
> Key: PHOENIX-1403
> URL: https://issues.apache.org/jira/browse/PHOENIX-1403
> Project: Phoenix
> Issue Type: Bug
> Affects Versions: 4.2, 3.2
> Reporter: Gabriel Reid
> Attachments: PHOENIX-1403.patch
>
>
> If there are multiple threads opening connections to the same Phoenix (HBase)
> instance concurrently, the same ConnectionQueryServicesImpl instance can be
> initialized multiple times. This can eventually cause some resources to leak,
> such as ZK connections.
> There is a synchronized block to prevent this, but PHOENIX-966 inadvertently
> disabled the function of the synchronized block.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
