[ 
https://issues.apache.org/jira/browse/SOLR-8983?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15246004#comment-15246004
 ] 

Varun Thacker commented on SOLR-8983:
-------------------------------------

Hi Anshum,

Awesome. The test now runs in 34s vs 55s ! Most of the time was spent waiting 
for the collection to come up live

{code}
6435 INFO  (qtp1682863192-26) [n:127.0.0.1:59727_solr    ] 
o.a.s.h.a.CollectionsHandler Wait for new collection to be active for at most 
30 seconds. Check all shard replicas
36507 ERROR (qtp1682863192-26) [n:127.0.0.1:59727_solr    ] 
o.a.s.h.a.CollectionsHandler Timed out waiting for new collection's replicas to 
become ACTIVE with timeout=30
{code}

Maybe since we know the collection is going to fail pass a custom solrXML 
string to MiniSolrCloudCluster which has  a smaller 
{{createCollectionWaitTimeTillActive}} ? 

bq. Is there a reason why you would want to pass that along? We aren't 
passing/appending the core responses and if the request was async, the user 
already received a response about request submission.

I guess adding the async param would make the call more resilient for longer 
delete times?

bq.  we don't even need to catch SolrException

I think we shouldn't change that . Sorry for the noise. It makes sense to catch 
it and throw it back so that the caller is informed correctly about a runtime 
exception.

Other changes look good! Thanks

> Failed Collection CREATE call should cleanup the cluster state before 
> returning
> -------------------------------------------------------------------------------
>
>                 Key: SOLR-8983
>                 URL: https://issues.apache.org/jira/browse/SOLR-8983
>             Project: Solr
>          Issue Type: Bug
>    Affects Versions: 5.5, 6.0
>            Reporter: Anshum Gupta
>            Assignee: Anshum Gupta
>             Fix For: 5.5.1, 6.1
>
>         Attachments: SOLR-8983-test.patch, SOLR-8983.patch, SOLR-8983.patch, 
> SOLR-8983.patch, SOLR-8983.patch
>
>
> In case of a failed collection creation call, the cluster state is updated 
> leaving an entry for the failed collection. This is also returned by the LIST 
> command, allowing the users to believe that the collection exists.
> CREATE call should cleanup in case of a failed attempt at creating the 
> collection.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

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

Reply via email to