cassandra-commits mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Jackson Chung (JIRA)" <j...@apache.org>
Subject [jira] [Commented] (CASSANDRA-7467) flood of "setting live ratio to maximum of 64" from repair
Date Sun, 29 Jun 2014 16:18:24 GMT

    [ https://issues.apache.org/jira/browse/CASSANDRA-7467?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14047150#comment-14047150
] 

Jackson Chung commented on CASSANDRA-7467:
------------------------------------------

please ignore prev comment. 2 nodes misbehaved over night while repair was running (on another
node), and crontab flush is already disabled.

> flood of "setting live ratio to maximum of 64" from repair
> ----------------------------------------------------------
>
>                 Key: CASSANDRA-7467
>                 URL: https://issues.apache.org/jira/browse/CASSANDRA-7467
>             Project: Cassandra
>          Issue Type: Bug
>            Reporter: Jackson Chung
>
> we are on 2.0.8
> running with repair -pr -local <KS>, all nodes on i2.2x (60G ram);, with setting
8G of heap. Using java 8. (key cache size is 1G)
> On occasion, when repair is run, the C* that run the repair, or another node in the cluster,
or both, run into a bad state with the system.log just printing ""setting live ratio to maximum
of 64"  forever every split seconds. It usually happens when repairing one of the larger/wider
CF. 
>  WARN [MemoryMeter:1] 2014-06-28 09:13:24,540 Memtable.java (line 470) setting live ratio
to maximum of 64.0 instead of Infinity
>  INFO [MemoryMeter:1] 2014-06-28 09:13:24,540 Memtable.java (line 481) CFS(Keyspace='RIQ',
ColumnFamily='MemberTimeline') liveRatio is 64.0 (just-counted was 64.0).  calculation took
0ms for 0 cells
> 		Table: MemberTimeline
> 		SSTable count: 13
> 		Space used (live), bytes: 17644018786
> ...
> 		Compacted partition minimum bytes: 30
> 		Compacted partition maximum bytes: 464228842
> 		Compacted partition mean bytes: 54578
> Just to give an idea of how bad this is, the log file is set to rotate 50 times with
21M each. In less than 15 minutes, all the logs are filled up with just that log. C* is not
responding, and can't be killed normally. Only way is to kill -9



--
This message was sent by Atlassian JIRA
(v6.2#6252)

Mime
View raw message