[ 
https://issues.apache.org/jira/browse/SOLR-16653?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17699701#comment-17699701
 ] 

Ishan Chattopadhyaya commented on SOLR-16653:
---------------------------------------------

Here's what I understand about the performance degradation and the fix.

Upon a collection creation, all replicas were created without "force-set-state" 
property on the replicas. Upon a restart, all collections were compared for its 
state info to see if force-set-state is "false", and forcefully write out the 
state.json with this "force-set-state" as false. All of this logic was 
introduced in https://issues.apache.org/jira/browse/SOLR-12087. Whereas this is 
not an additional overhead for non-PRS collections, this overhead can be 
reduced for PRS collections (and not require a persisting of state.json) by 
setting the flag to false to begin with, which is what this fix contains.

> Shard split doesn't work when shard has NRT and pull replica and prs is 
> enable on collection
> --------------------------------------------------------------------------------------------
>
>                 Key: SOLR-16653
>                 URL: https://issues.apache.org/jira/browse/SOLR-16653
>             Project: Solr
>          Issue Type: Bug
>      Security Level: Public(Default Security Level. Issues are Public) 
>          Components: SolrCloud
>    Affects Versions: 9.1.1
>            Reporter: Hitesh Khamesra
>            Assignee: Noble Paul
>            Priority: Blocker
>              Labels: pull-request-available
>             Fix For: 9.2
>
>         Attachments: Screenshot from 2023-02-25 06-41-17.png, Screenshot from 
> 2023-02-25 06-41-25.png, Screenshot from 2023-02-25 06-43-56.png, Screenshot 
> from 2023-02-27 12-17-06.png, Screenshot from 2023-02-28 12-59-01.png
>
>          Time Spent: 20m
>  Remaining Estimate: 0h
>
> To reproduce shard shard split scenario
>  # create a collection with PRS enable
>  # create shard with NRT and pull replica
>  # execute a shard split cmd
>  # result: subshard will not become active
>  
> Problem: We observed that prs state of replica is updated by data node. But 
> overseer node doesn't listen/read that updated prs state of replica. because 
> of that subshard remains inactive after shard split.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)

---------------------------------------------------------------------
To unsubscribe, e-mail: issues-unsubscr...@solr.apache.org
For additional commands, e-mail: issues-h...@solr.apache.org

Reply via email to