hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Stefan Will (JIRA)" <j...@apache.org>
Subject [jira] Commented: (HBASE-1988) OutOfMemoryError in RegionServer
Date Fri, 20 Nov 2009 00:32:39 GMT

    [ https://issues.apache.org/jira/browse/HBASE-1988?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12780347#action_12780347
] 

Stefan Will commented on HBASE-1988:
------------------------------------

Our HBase instance currently only has 700MB of data, and most scans are done by MR mappers
that all scan a very small amount of meta data (less than 10MB, I think). So the scans should
easily be able to be handled out of cache. At this point we're using HBase as a cache for
some precomputed data values, so the tables are neither particularly tall nor particularly
wide.

> OutOfMemoryError in RegionServer
> --------------------------------
>
>                 Key: HBASE-1988
>                 URL: https://issues.apache.org/jira/browse/HBASE-1988
>             Project: Hadoop HBase
>          Issue Type: Bug
>          Components: regionserver
>    Affects Versions: 0.20.2
>         Environment: Java 1.6.0_16 64bit on CentOS with 3000M max heap
>            Reporter: Stefan Will
>         Attachments: regionserver.log.gz
>
>
> RegionServers tend to die with an OutOfMemoryError under load. I expected this problem
to go away with the fix for HBASE-1927 in 0.20.2 RC1, but it's still happening. Also, when
this happens the cluster becomes unresponsive, even once the load on the machines has gone
back down. Interestingly there are lots and lots of scanner lease expired messages right before
the OOM.

-- 
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