Erick Erickson resolved SOLR-13091.
       Resolution: Fixed
    Fix Version/s: master (9.0)

Final patch attached.

The code changes for this fix in RebalanceLeaders isn't as extensive as they 
look, I did some refactoring.

The test file, OTOH, is pretty much rewritten. If we have similar breaks in the 
future I think the tests will catch them now.

This also fixes SOLR-10935.

I've changed the response. It should be more useful now. Probably tomorrow I'll 
fix SOLR-11998 and write up more of the documentation in the reference guide. 
The short form is:

1> I've added an overall "success" entry which gives the status of the actual 
leadership .vs. the preferredLeader. Before, the code manipulated the leader 
election queue and assumed that if those manipulations worked OK, then the 
leadership would actually change. This version tests that directly.

2> One question is what "success" is. In this case, success is returned if all 
the _active_ replicas that should be the leader are. So say you have a node 
that's down that has a replica on it that has the "preferredLeader" property 
set. "success" will be returned, along with an explanation.

When I fix SOLR-11998, probably tomorrow I'll also update the ref guide with 

> --------------------------
>                 Key: SOLR-13091
>                 URL: https://issues.apache.org/jira/browse/SOLR-13091
>             Project: Solr
>          Issue Type: Bug
>      Security Level: Public(Default Security Level. Issues are Public) 
>          Components: SolrCloud
>    Affects Versions: 7.6
>            Reporter: Erick Erickson
>            Assignee: Erick Erickson
>            Priority: Major
>             Fix For: 8.0, 7.7, master (9.0)
>         Attachments: SOLR-13091.patch, SOLR-13091.patch, SOLR-13091.patch
> From the user's list, the REBALANCELEADERS API doesn't seem to work 
> correctly. Manual testing shows the problem.
> It's disturbing that the unit test doesn't catch this. That'll be the first 
> thing to fix.

This message was sent by Atlassian JIRA

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

Reply via email to