[
https://issues.apache.org/jira/browse/DRILL-1519?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14166481#comment-14166481
]
Hanifi Gunes commented on DRILL-1519:
-------------------------------------
Sounds good. DRILL-1504 looks very much similar. My understanding is that the
description above is the root cause of the leak. I am not sure what happens
when the fragment memory limit is enabled/disabled though. Regardless, I expect
that the problem will persist.
> Concurrent access to WorkEventBus#getOrCreateFragmentManager leaks memory.
> --------------------------------------------------------------------------
>
> Key: DRILL-1519
> URL: https://issues.apache.org/jira/browse/DRILL-1519
> Project: Apache Drill
> Issue Type: Bug
> Reporter: Hanifi Gunes
> Assignee: Hanifi Gunes
>
> WorkEventBus uses a ConcurrentHashMap to ensure there is one and only one
> FragmentManager corresponding to each unique FragmentHandle. The method
> creates a FragmentManager that reserves some initial memory if a thread
> observes that it is not in the map. However, in case of concurrent access, we
> still need to clean up after if multiple threads observe that FragmentManager
> does not exist and create multiple FragmentManagers.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
