[
https://issues.apache.org/jira/browse/SOLR-18008?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18041739#comment-18041739
]
David Smiley commented on SOLR-18008:
-------------------------------------
Is the problem in any way reproducible -- like a script that would
create/delete in a loop? It'd be ideal to understand how this occurs.
>From my last job, I recall creating collections by the same name was
>problematic as Solr didn't have a good guarantee as to when exactly it would
>be able to create a collection by the same name again. Meaning Solr
>prematurely responded that it had deleted the collection, when it wasn't 100%
>done. So we changed our naming strategy to avoid the matter. I don't recall
>the symptom being about core data on disk, which is what you are reporting
>here, I thought it was a SolrCloud/ZK level matter. But I could be wrong.
> Allow Core Creation when remnant files are still on disk
> --------------------------------------------------------
>
> Key: SOLR-18008
> URL: https://issues.apache.org/jira/browse/SOLR-18008
> Project: Solr
> Issue Type: Bug
> Affects Versions: 8.11, 9.8
> Reporter: Eric Pugh
> Priority: Major
>
> Today in some production systems where we delete and re-create collections
> with the same name, we sometimes have left over files after deleting the the
> previous collection (and therefore the cores). The existence of these files
> causes any future collection creation to fail because when the individual
> core is created, we see we already have the remnant on disk.
> See the BATS test
> [https://github.com/apache/solr/blob/4d86c8cd128d59a65476dded34ec9c1947e67778/solr/packaging/test/test_create_collection2.bats]
> for an example of this situation.
> In a perfect world of course we wouldn't have these remnants, however I
> currently don't have the ability to debug why they intermittantly happen. I
> only know that when we go to make a new one we get errors similar to:
> Error CREATEing SolrCore xxxxxx: Could not create a new core in
> /var/solr/xxxxxx *as another core is already defined there*
> RequestHandlerBaseorg.apache.solr.common.SolrException: *Cannot unload
> non-existent core*
> Underlying core creation failed while creating collection
>
> The fix is unfortunantly to log onto the specific Solr box and execute a
> command similar to:
>
> rm -rf /var/solr/master_\{nameOfCollection}_\{indexMode}
> {shardNumber} \{replicaNumber}
> This Jira is to explore how we can at least handle this failure case in a
> better way, and allow the new create collection command to go through.
>
> I looked through old tickets, and in order of most likely related to least
> likely, here are some that I found:
> https://issues.apache.org/jira/browse/SOLR-7187
> https://issues.apache.org/jira/browse/SOLR-16792
> https://issues.apache.org/jira/browse/SOLR-17171
> https://issues.apache.org/jira/browse/SOLR-10599
> https://issues.apache.org/jira/browse/SOLR-9799
> https://issues.apache.org/jira/browse/SOLR-6598
> https://issues.apache.org/jira/browse/SOLR-14201
> https://issues.apache.org/jira/browse/SOLR-15674
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
