[hibernate-issues] [Hibernate-JIRA] Commented: (HSEARCH-557) Memory leak when using default ReaderProvider with Master/Slave Directories

[Sanne Grinovero (JIRA)]

    [ http://opensource.atlassian.com/projects/hibernate/browse/HSEARCH-557?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=37792#action_37792 ] 

Sanne Grinovero commented on HSEARCH-557:
-----------------------------------------

Hi, without reusing the same Directory instance you're introducing a memory leak: the default ReaderProvider is tracking open indexes using the Directory as a key, and for each Directory instance it will keep at least a single IndexReader open.
So in the HS implementation you'll have a minimum of 2 IndexReaders open (when all clients have closed all), but because you're closing Directories and opening new ones you're experiencing an unbounded number of open readers (each of which is caching the fields..).
This might explain why I'm not reproducing this; could you please check the behaviour on Hibernate Search 3.2 reverting your changes on the DirectoryProvider?

> Memory leak when using default ReaderProvider with Master/Slave Directories
> ---------------------------------------------------------------------------
>
>                 Key: HSEARCH-557
>                 URL: http://opensource.atlassian.com/projects/hibernate/browse/HSEARCH-557
>             Project: Hibernate Search
>          Issue Type: Bug
>          Components: directory provider
>    Affects Versions: 3.1.0.GA, 3.1.1.GA, 3.2.0.Final
>            Reporter: Sanne Grinovero
>             Fix For: 3.3.0
>
>
> The default ReaderProvider, org.hibernate.search.reader.SharingBufferReaderProvider is keeping a reference to an open index of the most current IndexReader, so it's able to refresh this on demand and track references to previously opened instances.
> The currently open indexReaders for the not-active directory might consume a lot of memory; follows a forum reference where it appears to be a memory leak; still not sure if it's a leak or just needing twice as much memory as otherwise, please comment here.
> https://forum.hibernate.org/viewtopic.php?f=9&t=1005540

-- 
This message is automatically generated by JIRA.
-
If you think it was sent incorrectly contact one of the administrators: http://opensource.atlassian.com/projects/hibernate/secure/Administrators.jspa
-
For more information on JIRA, see: http://www.atlassian.com/software/jira