[
https://issues.apache.org/jira/browse/CASSANDRA-1455?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12905971#action_12905971
]
Stu Hood commented on CASSANDRA-1455:
-------------------------------------
> host = range.endpoints.get(0)
Aha... good eye Flip. Looks like I broke that in 1322: so it is only collecting
the primary endpoints. That makes the "alternate through replicas of a range"
fix on 1434 moot, but you should still be seeing an even number of connections
to all the nodes in your cluster.
I'll roll another 1434, but I don't have any other ideas here at the moment: in
tests against a RF=1 12 node cluster (with OPP even), I saw an equal number of
connections to all nodes, even with the above bug.
> Hadoop streaming -> Cassandra ColumnFamilyOutputFormat not respecting
> partitioner
> ---------------------------------------------------------------------------------
>
> Key: CASSANDRA-1455
> URL: https://issues.apache.org/jira/browse/CASSANDRA-1455
> Project: Cassandra
> Issue Type: Bug
> Components: Hadoop
> Affects Versions: 0.7 beta 1
> Environment: Ubuntu lucid ; Hadoop 0.20.1 CDH3b1 ; Cassandra 0.7beta1
> + #1368, 1358, 1322, 1315 + patch to allow flat StreamingMutation avro schema
> Reporter: Philip (flip) Kromer
> Assignee: Stu Hood
> Priority: Critical
> Attachments: avromapper.rb, streamer.sh
>
>
> The Hadoop streaming shim (hadoop streaming client => avro =>
> ColumnFamilyOutputFormat => cassandra ring) is only connecting to one or a
> couple clients on the ring. With 24 hadoop clients launched, ` sudo netstat
> -antp | grep 9160 | wc -l ` gave 24 on one machine, and only 1-3 on any other
> node.
> I'll attach the script and runner I used.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
