[jira] [Commented] (SOLR-4709) dir lock error if reopening cores to fast?

["ASF subversion and git services (JIRA)" <ji...@apache.org>]

    [ https://issues.apache.org/jira/browse/SOLR-4709?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13829031#comment-13829031 ] 

ASF subversion and git services commented on SOLR-4709:
-------------------------------------------------------

Commit 1544220 from [~markrmiller@gmail.com] in branch 'dev/trunk'
[ https://svn.apache.org/r1544220 ]

SOLR-4709: The core reload after replication if config files have changed can fail due to a race condition.
SOLR-5489: TestIndexAndConfigAliasReplication commonly fails because it tries to get a lock for a locked index.
SOLR-5343: TestReplicationHandler.doTestStressReplication fails ~ 33% of the time

> dir lock error if reopening cores to fast?
> ------------------------------------------
>
>                 Key: SOLR-4709
>                 URL: https://issues.apache.org/jira/browse/SOLR-4709
>             Project: Solr
>          Issue Type: Bug
>            Reporter: Hoss Man
>            Assignee: Mark Miller
>             Fix For: 5.0, 4.7
>
>
> While testing my patch for SOLR-4629, i noticed a really random error that i traced back to the core reload (do to config file replication) failed because the directory was locked.
> From what i can tell, the lock checking code in the SolrCore constructor isn't even suppose to be used in the reload situation where there is a "prev" core, except that in SolrCore.reload there is this check...
> {noformat}
>     if (!getNewIndexDir().equals(getIndexDir())) {
>       // the directory is changing, don't pass on state
>       prev = null;
>     }
> {noformat}
> ..i'm not really sure i understand this logic, or what exactly the source of the problem is, or if the lock checking code should just be changed to work a differnet way completley, but it seemed worthy of tracking in it's own jira.
> log details to follow



--
This message was sent by Atlassian JIRA
(v6.1#6144)

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