lucene-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Robert Muir (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (LUCENE-3258) File leak when IOException occurs during index optimization.
Date Wed, 29 Jun 2011 19:14:28 GMT

    [ https://issues.apache.org/jira/browse/LUCENE-3258?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13057407#comment-13057407
] 

Robert Muir commented on LUCENE-3258:
-------------------------------------

just to followup, the changes here didn't make it until Lucene 3.3.0.
This isn't yet released, but should be any time soon (like within days)

you can try out the release candidate here: http://s.apache.org/lusolr330rc1

furthermore, if you want you can use lucene's test-framework jar in your own tests to help
you track down any file leaks in your own application, by wrapping your directory with MockDirectoryWrapper,
or by extending LuceneTestCase and using newDirectory() and newFSDirectory().

> File leak when IOException occurs during index optimization.
> ------------------------------------------------------------
>
>                 Key: LUCENE-3258
>                 URL: https://issues.apache.org/jira/browse/LUCENE-3258
>             Project: Lucene - Java
>          Issue Type: Bug
>          Components: core/index
>    Affects Versions: 3.0.3
>         Environment: SUSE Linux 11, Java 6
>            Reporter: Nick Kirsch
>
> I am not sure if this issue requires a fix due to the nature of its occurrence, or if
it exists in other versions of Lucene.
> I am using Lucene Java 3.0.3 on a SUSE Linux machine with Java 6 and have noticed there
are a number of file handles that are not being released from my java application. There are
IOExceptions in my log regarding disk full, which causes a merge and the optimization to fail.
The index is not currupt upon encountering the IOException. I am using CFS for my index format,
so 3X my largest index size during optimization certainly consumes all of my available disk.

> I realize that I need to add more disk space to my machine, but I investigated how to
clean up the leaking file handles. After failing to find a misuse of Lucene's IndexWriter
in the code I have wrapping Lucene, I did a quick search for close() being invoked in the
Lucene Jave source code. I found a number of source files that attempt to close more than
one object within the same close() method. I think a try/catch should be put around each of
these close() attempts to avoid skipping a subsequent closes. The catch may be able to ignore
a caught exception to avoid masking the original exception like done in SimpleFSDirectory.close().
> Locations in Lucene Java source where I suggest a try/catch should be used:
> - org.apache.lucene.index.FormatPostingFieldsWriter.finish()
> - org.apache.lucene.index.TermInfosWriter.close()
> - org.apache.lucene.index.SegmentTermPositions.close()
> - org.apache.lucene.index.SegmentMergeInfo.close()
> - org.apache.lucene.index.SegmentMerger.mergeTerms() (The finally block)
> - org.apache.lucene.index.DirectoryReader.close()
> - org.apache.lucene.index.FieldsReader.close()
> - org.apache.lucene.index.MultiLevelSkipListReader.close()
> - org.apache.lucene.index.MultipleTermPositions.close()
> - org.apache.lucene.index.SegmentMergeQueue.close()
> - org.apache.lucene.index.SegmentMergeDocs.close()
> - org.apache.lucene.index.TermInfosReader.close()

--
This message is automatically generated by JIRA.
For more information on JIRA, see: http://www.atlassian.com/software/jira

        

---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org


Mime
View raw message