hbase-issues mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Andrew Purtell (JIRA)" <j...@apache.org>
Subject [jira] Updated: (HBASE-2299) [EC2] mapreduce fixups for PE
Date Mon, 08 Mar 2010 09:45:27 GMT

     [ https://issues.apache.org/jira/browse/HBASE-2299?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]

Andrew Purtell updated HBASE-2299:
----------------------------------

    Attachment: HBASE-2299.patch

Attached patch addresses some issues:

- Update HADOOP_CLASSPATH in hadoop-env.sh to reference the current ZooKeeper jar packaged
with HBase.

- Do not enable mapper output compression by default. (DFS IO benchmark for one does not like
this.)

- Increase default task child heap to 512 MB.

- Add -XX:+UseCompressedOops to JVM flags globally. See http://wikis.sun.com/display/HotSpotInternals/CompressedOops
. On EC2 in just about all respects heap efficiency is the most important consideration. 


- Copy mapred configuration from Hadoop conf dir to HBase conf dir so launching mapreduce
jobs with 'hbase' script will do what is expected.

- Disable speculative execution.

- Try harder to insure HBase AMIs are found when looking for public images (just in case).


> [EC2] mapreduce fixups for PE
> -----------------------------
>
>                 Key: HBASE-2299
>                 URL: https://issues.apache.org/jira/browse/HBASE-2299
>             Project: Hadoop HBase
>          Issue Type: Sub-task
>            Reporter: Andrew Purtell
>            Assignee: Andrew Purtell
>             Fix For: 0.20.4, 0.21.0
>
>         Attachments: HBASE-2299.patch
>
>
> Scripts as packaged need tweaks for PE to successfully complete in multiclient (map-reduce)
mode. 

-- 
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