You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Shalin Shekhar Mangar (JIRA)" <ji...@apache.org> on 2017/07/12 09:40:00 UTC
[jira] [Commented] (SOLR-11045) The new replica created by
MoveReplica will have to have same name and coreName as the old one in case
of HDFS
[ https://issues.apache.org/jira/browse/SOLR-11045?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16083717#comment-16083717 ]
Shalin Shekhar Mangar commented on SOLR-11045:
----------------------------------------------
Thanks Dat. A few comments:
# minor nit -- the replica being checked for sharedFS in CloudUtil.checkSharedFSFailoverReplaced is different from the core descriptor being passed to the method. In reality it does not matter because today either all replicas are in shared FS or none but if it ever changes then this could be a tricky bug to find.
# In MoveReplicaHDFSTest, the {{assertNotSame(replica.getNodeName(), newReplica.getNodeName());}} is not correct. {{assertNotSame}} checks for object reference equality.
# The test should not extend MoveReplicaTest because then the super class's test() method is also executed as part of this test. Extend SolrCloudTestCase directly.
# The test should assertTrue on the result of {{ClusterStateUtil.waitForAllActiveAndLiveReplicas(zkStateReader, 120000);}} call otherwise a timeout is counted as success. On a local run I saw it hang for 2 minutes and yet the test succeeded.
> The new replica created by MoveReplica will have to have same name and coreName as the old one in case of HDFS
> --------------------------------------------------------------------------------------------------------------
>
> Key: SOLR-11045
> URL: https://issues.apache.org/jira/browse/SOLR-11045
> Project: Solr
> Issue Type: Bug
> Security Level: Public(Default Security Level. Issues are Public)
> Reporter: Cao Manh Dat
> Assignee: Cao Manh Dat
> Attachments: SOLR-11045.patch
>
>
> Currently, if the node contains a replica ( the replica being moved ) go down, then we call moveReplica command for it. That replica will only be removed from states.json. So If that node come back, we will have 2 replicas point to the same dataDir in HDFS.
> So if moveReplica can maintain the same name and coreName for the new replica, we can reuse {{CloudUtil.checkSharedFSFailoverReplaced()}} to clean the old one.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org