[
https://issues.apache.org/jira/browse/CASSANDRA-9610?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14608724#comment-14608724
]
Tyler Hobbs edited comment on CASSANDRA-9610 at 6/30/15 5:45 PM:
-----------------------------------------------------------------
Ah, okay, earlier you said there were 3 nodes in each DC in the 1.2.19 cluster.
The trace makes sense with 2 nodes in each DC.
I suspect that the problem in your 2.0.9 cluster is CASSANDRA-6930 and
CASSANDRA-7535. Basically, instead of combining all of queries for each token
range into one big query, they are being handled separately. This adds a lot
of overhead and can significantly slow the query down. I suggest seeing if you
can reproduce the issue on the latest 2.0.x version (currently 2.0.16).
was (Author: thobbs):
Ah, okay, earlier you said there were 3 nodes in each DC in the 1.2.19 cluster.
The trace makes sense with 2 nodes in each DC.
I suspect that the problem in your 2.0.9 cluster is CASSANDRA-6930. Basically,
instead of combining all of queries for each token range into one big query,
they are being handled separately. This adds a lot of overhead and can
significantly slow the query down. I suggest seeing if you can reproduce the
issue on the latest 2.0.x version (currently 2.0.16).
> Increased response time with cassandra 2.0.9 from 1.2.19
> --------------------------------------------------------
>
> Key: CASSANDRA-9610
> URL: https://issues.apache.org/jira/browse/CASSANDRA-9610
> Project: Cassandra
> Issue Type: Bug
> Reporter: Maitrayee
> Attachments: cluster_1.2.19, servicedefinition_schema.txt,
> traceout_1.2.19, traceout_2.0.9
>
>
> I was using Cassandra 1.2.19. Recently upgraded to 2.0.9. Queries with
> secondary index was completing much faster in 1.2.19
> Validated this with trace on via cqlsh
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
