cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Benedict (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-11026) OOM
Date Wed, 27 Jan 2016 15:04:40 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-11026?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15119393#comment-15119393
] 

Benedict commented on CASSANDRA-11026:
--------------------------------------

Prior to CASSANDRA-8099 the min/max buffers were "minimized" (i.e. they were extracted from
whatever enclosing buffer they were sourced from, into their own heap allocation).  This is
no longer the case, so the slabs they are allocated on get linked to, gradually consuming
heap as more sstables are flushed (a restart will bring temporary respite).

> OOM
> ---
>
>                 Key: CASSANDRA-11026
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11026
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Maxim Podkolzine
>         Attachments: Screenshot.png, dump.png
>
>
> Cassandra 3.0.2 fails with OOM. The heapdump shows large number of HeapByteBuffer instances,
each retaining 1Mb (see the details on the screenshot). Overall retained size is ~2Gb.
> We can provide the additional info and the whole heapdump if necessary.



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message