cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Paulo Motta (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-11344) Fix bloom filter sizing with LCS
Date Mon, 14 Mar 2016 23:06:33 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-11344?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15194350#comment-15194350
] 

Paulo Motta commented on CASSANDRA-11344:
-----------------------------------------

CASSANDRA-9830 results are much more consistent and bloom filter sizes have gone down from
60~100MB to ~10MB. The fix also is [confirmed|https://www.mail-archive.com/user@cassandra.apache.org/msg46633.html]
by a mailing user who have been facing OOMs due to large bloom filters.

I created a [regression dtest|https://github.com/pauloricardomg/cassandra-dtest/tree/11344]
to check that bloom filter are within a certain range (100K keys, bf should be between 50KB
and 100KB, it's 500KB in the broken version) and resubmitted above dtests with this branch.
Feel free to mark as ready to commit when/if tests are passing.

> Fix bloom filter sizing with LCS
> --------------------------------
>
>                 Key: CASSANDRA-11344
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-11344
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Marcus Eriksson
>            Assignee: Marcus Eriksson
>             Fix For: 2.2.x, 3.0.x, 3.x
>
>
> Since CASSANDRA-7272 we most often over allocate the bloom filter size with LCS



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

Mime
View raw message