[
https://issues.apache.org/jira/browse/NIFI-3631?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15961271#comment-15961271
]
ASF subversion and git services commented on NIFI-3631:
-------------------------------------------------------
Commit 778ba3957e7d1b9ddc105b1655f53e85cf6ec2ab in nifi's branch
refs/heads/master from [~markap14]
[ https://git-wip-us.apache.org/repos/asf?p=nifi.git;h=778ba39 ]
NIFI-3631: This closes #1613. Do not change Active Directory in
IndexDirectoryManager when it becomes full. Instead, wait until it is committed
and is removed by the onIndexCommitted method. This resolved a bug where the
index can exceed the configured limit but not yet be committed and as a result
would no longer be the active index. As a result, this bug causes the
IndexWriter never to get closed/removed from the IndexManager, and so a memory
leak is created
Signed-off-by: joewitt <[email protected]>
> Memory Leak in SimpleIndexManager
> ---------------------------------
>
> Key: NIFI-3631
> URL: https://issues.apache.org/jira/browse/NIFI-3631
> Project: Apache NiFi
> Issue Type: Bug
> Components: Core Framework
> Reporter: Mark Payne
> Assignee: Joseph Witt
> Priority: Critical
> Fix For: 1.2.0
>
>
> The SimpleIndexManager in the Provenance Repositories appears to leak
> instances of LuceneEventIndexWriter:
> [nifi@machine current]$ jmap -histo:live 31598 | grep LuceneEventIndexWriter
> 470: 1267 30408
> org.apache.nifi.provenance.lucene.LuceneEventIndexWriter$CommitStats
> 1266: 42 1680
> org.apache.nifi.provenance.lucene.LuceneEventIndexWriter
> We should have 1-2 Index Writers per prov partition, and this node has 2 prov
> partitions but we see 42 instances of the LuceneEventIndexWriter.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
