cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "daishi (JIRA)" <j...@apache.org>
Subject [jira] Commented: (CASSANDRA-157) make cassandra not allow itself to run out of memory during sustained inserts
Date Fri, 05 Jun 2009 01:44:07 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-157?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12716479#action_12716479
] 

daishi commented on CASSANDRA-157:
----------------------------------

> flushes are processed one at a time, first come first served, to avoid swamping the system
with IO.
> does that jive with what you saw? 

I guess that depends on what "first come first served" means.
Basically what I observed was that I could control how
frequently Memtable.flush() happened, but that only
after N memtables had been "queued up" to flush
would they actually be flushed to disk (I presume -
my view is based only on observing memory usage).
I think N was 10-20, but I don't remember exactly right now.


> make cassandra not allow itself to run out of memory during sustained inserts
> -----------------------------------------------------------------------------
>
>                 Key: CASSANDRA-157
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-157
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: daishi
>            Assignee: daishi
>         Attachments: Cassandra-157_Unregister_Memtable_MBean.diff
>
>
> Tv on IRC pointed out to me that the issue that I've been encountering
> is probably point 2. in this roadmap:
>     http://www.mail-archive.com/cassandra-dev@incubator.apache.org/msg00160.html
> I was unable to find any existing issue for this topic, so I'm creating a new one.
> Since this issue would block our use of Cassandra I'm happy to look into it,
> but if this is a known issue perhaps there's already a plan for addressing it
> that could be clarified?

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


Mime
View raw message