You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "David Smiley (JIRA)" <ji...@apache.org> on 2018/06/05 21:33:00 UTC
[jira] [Commented] (SOLR-11654)
TimePartitionedUpdateProcessor.lookupShardLeaderOfCollection should route
to the ideal shard
[ https://issues.apache.org/jira/browse/SOLR-11654?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16502530#comment-16502530 ]
David Smiley commented on SOLR-11654:
-------------------------------------
Looking at your patch. getLeaderNode I noticed doesn't pass along "slice" to the constructor of RetryNode (last param). It probably should do so for routeDocToSlice, but lookupShardLeadersOfCollections I guess null (what's happening today). I'm not completely clear on the implications yet.
Hey [~shalinmangar] or anyone who knows SolrCloud well, do you know if there are tests for ensuring the doc routing goes right to the correct shard leader instead of being arbitrary? I suppose it's fundamental to DURP but in CloudSolrClient AFAIK this is an optimization so perhaps there's a test for that? We're not sure how to test this since it's an internal routing thing. See Gus's most recent comment above. I directed Gus at TrackingShardHandlerFactory but didn't realize it's a search-only tracker not for updates (ouch).
> TimePartitionedUpdateProcessor.lookupShardLeaderOfCollection should route to the ideal shard
> --------------------------------------------------------------------------------------------
>
> Key: SOLR-11654
> URL: https://issues.apache.org/jira/browse/SOLR-11654
> Project: Solr
> Issue Type: Sub-task
> Security Level: Public(Default Security Level. Issues are Public)
> Components: SolrCloud
> Reporter: David Smiley
> Assignee: David Smiley
> Priority: Major
> Attachments: SOLR-11654.patch
>
>
> {{TimePartitionedUpdateProcessor.lookupShardLeaderOfCollection}} looks up the Shard/Slice to talk to for the given collection. It currently picks the first active Shard/Slice but it has a TODO to route to the ideal one based on the router configuration of the target collection. There is similar code in CloudSolrClient & DistributedUpdateProcessor that should probably be refactored/standardized so that we don't have to repeat this logic.
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org