[ https://issues.apache.org/jira/browse/HBASE-14574?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14947683#comment-14947683 ]
stack commented on HBASE-14574: ------------------------------- Hmm.... no... the reporting script is wrong. The zombie detector is right... I see them in the console output. > TableOutputFormat#getRecordWriter javadoc misleads > -------------------------------------------------- > > Key: HBASE-14574 > URL: https://issues.apache.org/jira/browse/HBASE-14574 > Project: HBase > Issue Type: Improvement > Reporter: stack > Assignee: stack > Attachments: 14574.txt > > > Some feedback from a user was that it was not plain that getRecordWriter > returns a NEW instance on each invocation. The base Interface javadoc on > OutputWriter would seem to indicate there is one instance for a whole > 'job'... That is not what we do. So, just reading the javadoc, you'd not > bother with closing the returned record writer since the framework will do > the close up for you; thats what the javadoc implies. This made it so it > seemed hbase was losing writes. > Lets fix in javadoc at least. -- This message was sent by Atlassian JIRA (v6.3.4#6332)