[ 
https://issues.apache.org/jira/browse/HADOOP-9455?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=13628194#comment-13628194
 ] 

Chris Nauroth commented on HADOOP-9455:
---------------------------------------

Eli, thank you for taking a look.

{quote}
Wouldn't it be better to only append HADOOP_CLIENT_OPTS once in hadoop-env.sh 
rather than in each per-project bin script?
{quote}

This would cause a subtle change in behavior.  Right now, the hdfs script only 
appends HADOOP_CLIENT_OPTS for a subset of the interactive commands: dfs, 
dfsadmin, haadmin, and fsck.  This is not done for the daemon commands: 
namenode, datanode, secondarynamenode, etc.  It appears that the intent is that 
when you want to override options to a daemon, you use the daemon-specific 
environment variable instead: HADOOP_NAMENODE_OPTS, HADOOP_DATANODE_OPTS, 
HADOOP_SECONDARYNAMENODE_OPTS, etc.  It could be erroneous to combine options 
set in HADOOP_CLIENT_OPTS with options set for a daemon.  For example, I could 
imagine an operator wanting to maintain a single hadoop-env.sh that sets both 
HADOOP_NAMENODE_OPTS and HADOOP_CLIENT_OPTS, but set different values for max 
heap size within each one.

                
> HADOOP_CLIENT_OPTS is appended twice, causing JVM failures
> ----------------------------------------------------------
>
>                 Key: HADOOP-9455
>                 URL: https://issues.apache.org/jira/browse/HADOOP-9455
>             Project: Hadoop Common
>          Issue Type: Bug
>          Components: bin
>    Affects Versions: 3.0.0, 2.0.3-alpha
>            Reporter: Sangjin Lee
>            Assignee: Chris Nauroth
>            Priority: Minor
>         Attachments: HADOOP-9455.1.patch
>
>
> If you set HADOOP_CLIENT_OPTS and run hadoop, you'll find that the 
> HADOOP_CLIENT_OPTS value gets appended twice, and leads to JVM start failures 
> for cases like adding debug flags.
> For example,
> {noformat}
> HADOOP_CLIENT_OPTS='-agentlib:jdwp=transport=dt_socket,address=localhost:9009,server=y,suspend=y'
>  hadoop jar anything
> ERROR: Cannot load this JVM TI agent twice, check your java command line for 
> duplicate jdwp options.
> Error occurred during initialization of VM
> agent library failed to init: jdwp
> {noformat}

--
This message is automatically generated by JIRA.
If you think it was sent incorrectly, please contact your JIRA administrators
For more information on JIRA, see: http://www.atlassian.com/software/jira

Reply via email to