You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Cao Manh Dat (JIRA)" <ji...@apache.org> on 2015/12/30 09:01:49 UTC
[jira] [Comment Edited] (SOLR-8461) CloudSolrStream and
ParallelStream can choose replicas that are not active
[ https://issues.apache.org/jira/browse/SOLR-8461?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15074759#comment-15074759 ]
Cao Manh Dat edited comment on SOLR-8461 at 12/30/15 8:01 AM:
--------------------------------------------------------------
{quote}
1. With the patch we check if the replica is active, we should also also check if the replica's nodeName belongs in the live nodes list. This is required for a scenario like this - Someone kills the node using "kill -9 <pid>" / OOM crash . In the cluster state that replica will still show us "active". HttpSolrCall#getCoreByCollection does the same thing
{quote}
Very nice suggestion, I Will update it now.
{quote}
2. Maybe move the code into a function - say "getActiveReplicasForCollection" ? The both CloudSolrStream and ParallelSteam can reuse it.
{quote}
Cant do that. In ParallelStream, we just wanna find enough workers, dont care about what collection/shard that replica belong to. In CloudSolrStream, for each collection we will randomly find a replica per shard.
was (Author: caomanhdat):
{{quote}}
1. With the patch we check if the replica is active, we should also also check if the replica's nodeName belongs in the live nodes list. This is required for a scenario like this - Someone kills the node using "kill -9 <pid>" / OOM crash . In the cluster state that replica will still show us "active". HttpSolrCall#getCoreByCollection does the same thing
{{quote}}
Very nice suggestion, I Will update it now.
{{quote}}
2. Maybe move the code into a function - say "getActiveReplicasForCollection" ? The both CloudSolrStream and ParallelSteam can reuse it.
{{quote}}
Cant do that. In ParallelStream, we just wanna find enough workers, dont care about what collection/shard that replica belong to. In CloudSolrStream, for each collection we will randomly find a replica per shard.
> CloudSolrStream and ParallelStream can choose replicas that are not active
> --------------------------------------------------------------------------
>
> Key: SOLR-8461
> URL: https://issues.apache.org/jira/browse/SOLR-8461
> Project: Solr
> Issue Type: Bug
> Reporter: Joel Bernstein
> Attachments: SOLR-8461.patch, SOLR-8461.patch, SOLR-8461.patch
>
>
> Currently CloudSolrStream and ParallelStream don't check the state of the replicas they route requests to. This can result in replicas that are not active receiving request.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org