[ https://issues.apache.org/jira/browse/HIVE-8914?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14219900#comment-14219900 ]
Eugene Koifman commented on HIVE-8914: -------------------------------------- I see that new FileSystem is created but never closed so I don't see how this addresses the issue. > HDFSCleanup thread holds reference to FileSystem > ------------------------------------------------ > > Key: HIVE-8914 > URL: https://issues.apache.org/jira/browse/HIVE-8914 > Project: Hive > Issue Type: Bug > Components: WebHCat > Affects Versions: 0.13.1 > Reporter: shanyu zhao > Assignee: shanyu zhao > Attachments: HIVE-8914.patch > > > WebHCat server has a long running cleanup thread (HDFSCleanup) which holds a > reference to FileSystem. Because of this reference, many FileSystem related > objects (e.g. MetricsSystemImpl) cannot be garbage collected. Sometimes this > causes OOM exception. Since the cleanup is done every 12 hours by default, we > can simply recreate a FileSystem when we need to use it. -- This message was sent by Atlassian JIRA (v6.3.4#6332)