You are viewing a plain text version of this content. The canonical link for it is here.
Posted to solr-dev@lucene.apache.org by "Shalin Shekhar Mangar (JIRA)" <ji...@apache.org> on 2009/01/24 08:45:59 UTC
[jira] Updated: (SOLR-978) Old files are not removed from slaves
after replication
[ https://issues.apache.org/jira/browse/SOLR-978?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Shalin Shekhar Mangar updated SOLR-978:
---------------------------------------
Attachment: SOLR-978.patch
Changes
# Add a forceOpenWriter method to DirectUpdateHandler2 which is called by SnapPuller on the slave to make sure IndexWriter is open. This is to make sure IndexDeletionPolicy gets a chance to cleanup older commit points.
# SnapPuller calls forceOpenWriter if core.getUpdateHandler is an instance of DirectUpdateHandler2.
# SnapPuller calls delTree on temporary files and folders in a finally block
I know 2 is not a very clean way. I thought of adding that method in UpdateHandler as a no-op and overriding it in DUH2. However that would mean that custom UpdateHandler implementations would silently ignore this call and we will not have a chance to warn them (which we can do with this patch). Also, DUH2 already has a protected openWriter method. However calling that method needs access to a protected lock object.
> Old files are not removed from slaves after replication
> -------------------------------------------------------
>
> Key: SOLR-978
> URL: https://issues.apache.org/jira/browse/SOLR-978
> Project: Solr
> Issue Type: Bug
> Components: replication (java)
> Reporter: Noble Paul
> Assignee: Shalin Shekhar Mangar
> Fix For: 1.4
>
> Attachments: SOLR-978.patch
>
>
> The files in the slave's index is not removed even after they are removed from master.
> The temp directories named index.xxxxxxx are not removed after use
> The problem is that when we do a commit on the slave after replication is done. The commit does not re-open the IndexWriter. Therefore, the deletion policy does not take affect and older files are left as is. This can keep on building up. The only solution is to re-open the index writer.
> A user has reported the issue http://markmail.org/thread/yw5n4dk2t5zbt5z5
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.