[ 
https://issues.apache.org/jira/browse/SOLR-11665?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16523548#comment-16523548
 ] 

Andrzej Bialecki  commented on SOLR-11665:
------------------------------------------

This patch contains the following improvements and fixes:
 * failures at any phase in the process are followed by an explicit cleanup, 
which removes partially constructed shards
 * sub-shards are now created with the same number of replicas per replica type 
as the parent type - previously only NRT replicas would be created.
 * fixed a bug in coreName construction so that the new cores follow the same 
naming pattern per replica type as the parent shards
 * SplitShardCmd now first checks the amount of available disk space on the 
parent shard leader's node to ensure that the actual index splitting has enough 
disk space to proceed.
 * unit test changes, and a test that verifies the correct number of replicas 
per type in the new sub-shards.

> Improve SplitShardCmd error handling
> ------------------------------------
>
>                 Key: SOLR-11665
>                 URL: https://issues.apache.org/jira/browse/SOLR-11665
>             Project: Solr
>          Issue Type: Bug
>      Security Level: Public(Default Security Level. Issues are Public) 
>          Components: AutoScaling
>            Reporter: Cao Manh Dat
>            Assignee: Andrzej Bialecki 
>            Priority: Major
>         Attachments: SOLR-11665.patch, SOLR-11665.test.patch
>
>
> I do see some problems when doing split shard but there are no available 
> nodes for creating replicas ( due to policy framework )
> - The patch contains a test, in which sub shard stay in CONSTRUCTION state 
> forever.
> - Shard which get split, stay in inactive state forever and subshards are not 
> created



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)

---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]

Reply via email to