You are viewing a plain text version of this content. The canonical link for it is here.
Posted to notifications@accumulo.apache.org by "Josh Elser (JIRA)" <ji...@apache.org> on 2014/01/11 00:38:50 UTC
[jira] [Updated] (ACCUMULO-1493) documentation clarification for
MapR
[ https://issues.apache.org/jira/browse/ACCUMULO-1493?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]
Josh Elser updated ACCUMULO-1493:
---------------------------------
Fix Version/s: (was: 1.5.1)
(was: 1.6.0)
> documentation clarification for MapR
> ------------------------------------
>
> Key: ACCUMULO-1493
> URL: https://issues.apache.org/jira/browse/ACCUMULO-1493
> Project: Accumulo
> Issue Type: Improvement
> Components: docs
> Affects Versions: 1.5.0
> Environment: MapR 2.1.1
> Reporter: Keys Botzum
> Priority: Minor
> Labels: Documentation
>
> When testing Accumulo with MapR we encountered a few minor issues that should be addressable via documentation. They are:
> 1) Accumulo 1.5 requires these jar files:
> commons-configuration-1.6.jar
> commons-collections-3.2.1.jar
> commons-io-2.1.jar
> They are not part of the MapR distribution and do not appear to be part of the standard Hadoop distribution either. These dependencies should be documented.
> 2) MapR has the requirement that the block/chunk size be a multiple of 64K. As a result the default Accumulo algorithm for defining the WAL block size does not work with MapR. This should be documented. In more detail...
> By default, the property tserver.wal.blocksize is 0, in which case Accumlo sets it to 1.1 * tserver.walog.max.size, which defaults to 512M. This results in a MapR chunksize that is not a multiple of 64k, which means any writes to the WAL will fail. So in accumulo-site.xml, set tserver.wal.blocksize to a multiple of 64K (such as 562M, which is about 1.1 times the WAL max. Insert these lines:
> <property>
> <name>tserver.wal.blocksize</name>
> <value>562M</value>
> </property>
> 3) Need to document the Accumulo log sync mechanism for MapR. Accumulo 1.5 changed the way the WAL is sync'ed and now uses the distributed file system. A filesystem-specific property is needed to specify the class for MapR. Add this to accumulo-site.xml:
> <property>
> <name>master.walog.closer.implementation</name>
> <value>org.apache.accumulo.server.master.recovery.MapRLogCloser</value>
> </property>
--
This message was sent by Atlassian JIRA
(v6.1.5#6160)