[ 
https://issues.apache.org/jira/browse/CASSANDRA-1461?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12906062#action_12906062
 ] 

Jeremy Hanna commented on CASSANDRA-1461:
-----------------------------------------

it might also be nice to make these settings update-able via JMX.  That way, if 
the server is bogged down, an admin could disable rows_cached or limit the 
number of keys_cached.  Both would need to be immediately update-able.  So it 
would need to probably update the global capacity value for those caches to 
appropriate values - 0 for rows_cached, some other limit for keys_cached.  That 
way the actual cache data structure could be reduced and the memory could be 
reclaimed.

> Make rows_cached and keys_cached globally override-able
> -------------------------------------------------------
>
>                 Key: CASSANDRA-1461
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-1461
>             Project: Cassandra
>          Issue Type: Improvement
>          Components: Core
>            Reporter: Jeremy Hanna
>            Assignee: Jeremy Hanna
>
> It would be nice to make rows_cached and keys_cached globally override-able.
> rows_cached - options to disable globally
> It would be nice to have a setting to be able to globally disable the 
> rows_cached setting.  In a system where there are multiple users/applications 
> on a single cassandra cluster, rows_cached can become unmanageable in its 
> current form.  It's just too memory intensive and with competing applications 
> on a single cluster, it's hard to guarantee isolation and QoS.
> keys_cached - option to have a global limit override
> When there are multiple users/application, even the keys_cached could 
> potentially get out of control.  So it would be nice to be able give an 
> overall limit on how many keys can be cached per column family or possibly a 
> total for the keyspace, as keyspace usually translates into a specific 
> application.  So once the column family has reached the global limit, it 
> would just stay at that limit.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to