You are viewing a plain text version of this content. The canonical link for it is here.

Posted to dev@lucene.apache.org by "ASF subversion and git services (JIRA)" <ji...@apache.org> on 2014/08/21 17:05:12 UTC

[jira] [Commented] (SOLR-6089) When using the HDFS block cache, when a file is deleted, it's underlying data entries in the block cache are not removed, which is a problem with the global block cache option.

    [ https://issues.apache.org/jira/browse/SOLR-6089?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14105441#comment-14105441 ] 

ASF subversion and git services commented on SOLR-6089:
-------------------------------------------------------

Commit 1619427 from [~markrmiller@gmail.com] in branch 'dev/trunk'
[ https://svn.apache.org/r1619427 ]

SOLR-6089: When using the HDFS block cache, when a file is deleted, it's underlying data entries in the block cache are not removed, which is a problem with the global block cache option.

> When using the HDFS block cache, when a file is deleted, it's underlying data entries in the block cache are not removed, which is a problem with the global block cache option.
> --------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
>
>                 Key: SOLR-6089
>                 URL: https://issues.apache.org/jira/browse/SOLR-6089
>             Project: Solr
>          Issue Type: Bug
>          Components: hdfs
>            Reporter: Mark Miller
>            Assignee: Mark Miller
>         Attachments: SOLR-6089.patch
>
>
> Patrick Hunt noticed this. Without the global block cache, the block cache was not reused after a directory was closed. Now that it is reused when using the global cache, leaving the underlying entries presents a problem if that directory is created again because blocks from the previous directory may be read. This could happen when you remove a solrcore and recreate it with the same data directory (or a collection with the same name). I could only reproduce it easily using index merges (core admin) with the sequence: merge index, delete collection, create collection, merge index. Reads on the final merged index can look corrupt or queries may just return no results.



--
This message was sent by Atlassian JIRA
(v6.2#6252)

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