You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@lucene.apache.org by "Yibo Hu (JIRA)" <ji...@apache.org> on 2017/11/13 03:37:00 UTC
[jira] [Updated] (SOLR-11639) Improve the LBHttpSolrClient to aware
of the load change faster
[ https://issues.apache.org/jira/browse/SOLR-11639?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Yibo Hu updated SOLR-11639:
---------------------------
Description:
The LBHttpSolrClient is referenced by HttpShardHandlerFactory.
In my solrcloud, I config all collection with 2 replicas.
When one node of the solrcloud was stopped by a long time gc ,before the liveNodes of zookeeper drop it , half of the request would be timeout.
Should we add a counter of connection count of all nodes?
If one node is stopped by gc(The node is not dead ),the connection count of this node would be larger than the other node. So most of the requests will lead to the low connection count node.
was:
The LBHttpSolrClient is referenced by HttpShardHandlerFactory.
In my solrcloud, I config all collection with 2 replicas.
When one node of the solrcloud was stopped by a long time gc ,before the liveNodes of zookeeper drop it , half of the request would be timeout.
Should we add a counter of connection count of all nodes?
If one node is stopped by gc(The node is not dead ),the connection count of this node would be larger then the other node. So most of the requests will lead to the low connection count node.
> Improve the LBHttpSolrClient to aware of the load change faster
> ---------------------------------------------------------------
>
> Key: SOLR-11639
> URL: https://issues.apache.org/jira/browse/SOLR-11639
> Project: Solr
> Issue Type: Improvement
> Security Level: Public(Default Security Level. Issues are Public)
> Components: clients - java
> Reporter: Yibo Hu
> Labels: performance
>
> The LBHttpSolrClient is referenced by HttpShardHandlerFactory.
> In my solrcloud, I config all collection with 2 replicas.
> When one node of the solrcloud was stopped by a long time gc ,before the liveNodes of zookeeper drop it , half of the request would be timeout.
> Should we add a counter of connection count of all nodes?
> If one node is stopped by gc(The node is not dead ),the connection count of this node would be larger than the other node. So most of the requests will lead to the low connection count node.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: dev-unsubscribe@lucene.apache.org
For additional commands, e-mail: dev-help@lucene.apache.org