[ 
https://issues.apache.org/jira/browse/HADOOP-2766?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=12564801#action_12564801
 ] 

Karam Singh commented on HADOOP-2766:
-------------------------------------

+1 
Ran hod after applying patch with follwoing options -:
hod -c conf/hodrc0.16 -b 4 -o "allocate clusterDir1 3" 
--gridservice-mapred.envs='HADOOP_OPTS=-Dmy.env\=\"Hello\,World\:Test\/test1\;tt\"\
 -Xms512m\ -XX\:\-UseParallelGC 
-XX\:\-RelaxAccessControlCheck,HADOOP_HEAPSIZE=2048,HADOOP_ROOT_LOGGER=DEBUG\,console'
 --gridservice-hdfs.envs="HADOOP_OPTS=-Dmy.env\=\'World\:Test\'\ 
-Dmy.env2\=\'hodtest/HOD\'\ -XX\:\+MaxFDLimit -
XX\:\+UseBoundThreads 
-XX\:\+UseTLAB,HADOOP_CLASSPATH=/home\:/proc,HADOOP_HEAPSIZE=2048,HADOOP_ROOT_LOGGER=DEBUG\,console"

Also tried it by setting value in hodrc.
Note--: If in hodrc some value of key=value params contains ';' then value from 
';' afterwards will be ignored. Same thing works from commandline.
Tried with bot hadoop 0.16 and hadoop 0.15 
Also ran smoke test successfully with static dfs 

> [HOD] No way to set HADOOP_OPTS environment variable to the Hadoop daemons 
> through HOD
> --------------------------------------------------------------------------------------
>
>                 Key: HADOOP-2766
>                 URL: https://issues.apache.org/jira/browse/HADOOP-2766
>             Project: Hadoop Core
>          Issue Type: Bug
>          Components: contrib/hod
>    Affects Versions: 0.16.0
>            Reporter: Hemanth Yamijala
>            Assignee: Vinod Kumar Vavilapalli
>            Priority: Critical
>         Attachments: hadoop-2766
>
>
> For purposes of performance tuning it should be possible to set some 
> environment variables that Hadoop honors before launching the Hadoop daemons. 
> For e.g. to set the heap size of the JVM, we should set the HADOOP_HEAPSIZE 
> variable. These can be configured through hodrc using the 
> gridservice-mapred.envs environment variable. This works for everything 
> *except* HADOOP_OPTS, which is also useful, for e.g. to pass in Garbage 
> collection parameters to the JVM, like -XX:+UseParallelGC. This is because 
> HOD tries to set HADOOP_OPTS from the gridservice-mapred.java-opts variable, 
> but it does not read the java-opts variable anywhere properly.

-- 
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.

Reply via email to