[
https://issues.apache.org/jira/browse/HIVE-13151?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15205392#comment-15205392
]
Eugene Koifman commented on HIVE-13151:
---------------------------------------
Shouldn't this be solved in Hadoop? If anything else is using UGI the cache
will still leak. What do other submodules do?
1. TestTxnCommands2 - seems to have unused imports added
2. Would it make sense to include CompactionInfo.getFullPartitionName() in the
error message for more context? (Worker/Initiator/Cleaner - basically
throughout)
> Clean up UGI objects in FileSystem cache for transactions
> ---------------------------------------------------------
>
> Key: HIVE-13151
> URL: https://issues.apache.org/jira/browse/HIVE-13151
> Project: Hive
> Issue Type: Bug
> Components: Hive
> Affects Versions: 2.0.0
> Reporter: Wei Zheng
> Assignee: Wei Zheng
> Attachments: HIVE-13151.1.patch, HIVE-13151.2.patch,
> HIVE-13151.3.patch
>
>
> One issue with FileSystem.CACHE is that it does not clean itself. The key in
> that cache includes UGI object. When new UGI objects are created and used
> with the FileSystem api, new entries get added to the cache.
> We need to manually clean up those UGI objects once they are no longer in use.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
