[
https://issues.apache.org/jira/browse/SOLR-13924?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16972864#comment-16972864
]
Shalin Shekhar Mangar commented on SOLR-13924:
----------------------------------------------
The changes to the MoveReplicaTest in SOLR-13843 started adding either a TLOG
or a PULL replica to the collection in addition to the NRT replica. This
exposed a genuine bug. Move replica operation does not work for PULL replicas
stored on HDFS. We just didn't know so far because there were no tests for it.
Basically move replica operation for HDFS assumes that each replica has an
update log directory containing the tlog files but the PULL type replicas do
not have them. This causes the NPE.
> MoveReplica failures when using HDFS (NullPointerException)
> -----------------------------------------------------------
>
> Key: SOLR-13924
> URL: https://issues.apache.org/jira/browse/SOLR-13924
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Affects Versions: 8.3
> Reporter: Chris M. Hostetter
> Assignee: Shalin Shekhar Mangar
> Priority: Major
>
> Based on recent jenkins test failures, it appears that attemping to use the
> "MoveReplica" command on HDFS has a high chance of failure due to an
> underlying NPE.
> I'm not sure if this bug *only* affects HDFS, or if it's just more likly to
> occur when using HDFS due to some timing quirks.
> It's also possible that the bug impacts non-HDFS users just as much as HDFS
> users, but only manifests in our tests due to some quick of our
> {{cloud-hdfs}} test configs.
> The problem appears to be new in 8.3 as a result of changes made in SOLR-13843
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@lucene.apache.org
For additional commands, e-mail: issues-h...@lucene.apache.org
