[jira] Updated: (CASSANDRA-2148) system CFs default to large memtable throughputs on large heaps

["Matthew F. Dennis (JIRA)" <ji...@apache.org>]

     [ https://issues.apache.org/jira/browse/CASSANDRA-2148?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Matthew F. Dennis updated CASSANDRA-2148:
-----------------------------------------

    Attachment: 2148-cassandra-0.7.txt

attached patch bounds Hints CF throughput between 32M and 256M and fixes the throughput of other system CFs to 8M

> system CFs default to large memtable throughputs on large heaps
> ---------------------------------------------------------------
>
>                 Key: CASSANDRA-2148
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-2148
>             Project: Cassandra
>          Issue Type: Bug
>    Affects Versions: 0.7.0
>            Reporter: Matthew F. Dennis
>            Assignee: Matthew F. Dennis
>             Fix For: 0.7.2
>
>         Attachments: 2148-cassandra-0.7.txt
>
>
> The default memtableThroughputInMB is calculated now based on the heap size.  Most people running with a large heap in production explicitly set it for their memtable(s).  However, the the CFs in the system keyspace still default to the calculated value which on a large heap can be quite large.  HintsColumnFamily is really the only problematic one though as the others are flushed afters changes to them.
> we should:
> 1) set the throughput on the hints CF to a reasonable max and min value - min(256, max(32, normalDefault/2))
> 2) set the throughput on the other system CFs to some small constant value (just as a safety); 8M sounds good

-- 
This message is automatically generated by JIRA.
-
For more information on JIRA, see: http://www.atlassian.com/software/jira