[
https://issues.apache.org/jira/browse/CASSANDRA-11737?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15276554#comment-15276554
]
Jeremiah Jordan edited comment on CASSANDRA-11737 at 5/9/16 4:06 PM:
---------------------------------------------------------------------
https://github.com/JeremiahDJordan/cassandra/commits/CASSANDRA-11737-21
That merges up cleanly all the way to trunk. Adds a -D option to disable using
severity in DynamicEndpointSnitch calculations.
For the clusters where we were seeing massive latency degredation when a single
node was under load applying this patch and settings the -D to disable severity
brought things back to normal.
was (Author: jjordan):
https://github.com/JeremiahDJordan/cassandra/commits/CASSANDRA-11737-21
That merges up cleanly all the way to trunk. Adds a -D option to disable using
severity in DynamicEndpointSnitch calculations.
> Add a way to disable severity in DynamicEndpointSnitch
> ------------------------------------------------------
>
> Key: CASSANDRA-11737
> URL: https://issues.apache.org/jira/browse/CASSANDRA-11737
> Project: Cassandra
> Issue Type: Bug
> Reporter: Jeremiah Jordan
> Assignee: Jeremiah Jordan
>
> I have seen in a few clusters now where severity can out weigh latency in
> DynamicEndpointSnitch causing issues (a node that is completely overloaded
> CPU wise and has super high latency will get selected for queries, even
> though nodes with much lower latency exist, but they have a higher severity
> score). There should be a way to disable the use of severity.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
