You are viewing a plain text version of this content. The canonical link for it is here.
Posted to common-issues@hadoop.apache.org by "Jonathan Ellithorpe (JIRA)" <ji...@apache.org> on 2015/11/17 06:15:10 UTC

[jira] [Updated] (HADOOP-12574) Multiple Xmx Parameters to Java Process

     [ https://issues.apache.org/jira/browse/HADOOP-12574?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Jonathan Ellithorpe updated HADOOP-12574:
-----------------------------------------
    Description: 
When launching hadoop, the setting of environment variable HADOOP_HEAPMAX results in multiple Xmx parameters being passed to the java application launcher. This is because a default Xmx setting included in the HADOOP_CLIENT_OPTS in etc/hadoop/hadoop-env.sh, line 61:

{noformat}
# The following applies to multiple commands (fs, dfs, fsck, distcp etc)
export HADOOP_CLIENT_OPTS="-Xmx512m $HADOOP_CLIENT_OPTS"
{noformat}

is merged into the variable HADOOP_OPTS in bin/hadoop, line 160:

{noformat}
    # Always respect HADOOP_OPTS and HADOOP_CLIENT_OPTS
    HADOOP_OPTS="$HADOOP_OPTS $HADOOP_CLIENT_OPTS"
{noformat}

Orthogonally, the HADOOP_HEAPMAX environment variable triggers the setting of JAVA_HEAP_MAX in hadoop-config.sh, line 192:

{noformat}
# check envvars which might override default args
if [ "$HADOOP_HEAPSIZE" != "" ]; then
  #echo "run with heapsize $HADOOP_HEAPSIZE"
  JAVA_HEAP_MAX="-Xmx""$HADOOP_HEAPSIZE""m"
  #echo $JAVA_HEAP_MAX
fi
{noformat}

Which is ultimately passed to the java application launcher in bin/hadoop, line 166:

{noformat}
    exec "$JAVA" $JAVA_HEAP_MAX $HADOOP_OPTS $CLASS "$@"
{noformat}

The result: two Xmx settings passed to the Java application launcher. Which one is respected depends on the JVM, but for HotSpot the rightmost setting is respected, and so the HADOOP_HEAPMAX environment variable has no effect and the default setting in HADOOP_CLIENT_OPTS wins. 

  was:
When launching hadoop, the setting of environment variable HADOOP_HEAPMAX results in multiple Xmx parameters being passed to the java application launcher. This is because a default Xmx setting included in the HADOOP_CLIENT_OPTS in etc/hadoop/hadoop-env.sh, line 61:

# The following applies to multiple commands (fs, dfs, fsck, distcp etc)
export HADOOP_CLIENT_OPTS="-Xmx512m $HADOOP_CLIENT_OPTS"

is merged into the variable HADOOP_OPTS in bin/hadoop, line 160:

    # Always respect HADOOP_OPTS and HADOOP_CLIENT_OPTS
    HADOOP_OPTS="$HADOOP_OPTS $HADOOP_CLIENT_OPTS"

Orthogonally, the HADOOP_HEAPMAX environment variable triggers the setting of JAVA_HEAP_MAX in hadoop-config.sh, line 192:

# check envvars which might override default args
if [ "$HADOOP_HEAPSIZE" != "" ]; then
  #echo "run with heapsize $HADOOP_HEAPSIZE"
  JAVA_HEAP_MAX="-Xmx""$HADOOP_HEAPSIZE""m"
  #echo $JAVA_HEAP_MAX
fi

Which is ultimately passed to the java application launcher in bin/hadoop, line 166:

    exec "$JAVA" $JAVA_HEAP_MAX $HADOOP_OPTS $CLASS "$@"

The result: two Xmx settings passed to the Java application launcher. Which one is respected depends on the JVM, but for HotSpot the rightmost setting is respected, and so the HADOOP_HEAPMAX environment variable has no effect and the default setting in HADOOP_CLIENT_OPTS wins. 


> Multiple Xmx Parameters to Java Process
> ---------------------------------------
>
>                 Key: HADOOP-12574
>                 URL: https://issues.apache.org/jira/browse/HADOOP-12574
>             Project: Hadoop Common
>          Issue Type: Bug
>    Affects Versions: 2.7.1
>            Reporter: Jonathan Ellithorpe
>
> When launching hadoop, the setting of environment variable HADOOP_HEAPMAX results in multiple Xmx parameters being passed to the java application launcher. This is because a default Xmx setting included in the HADOOP_CLIENT_OPTS in etc/hadoop/hadoop-env.sh, line 61:
> {noformat}
> # The following applies to multiple commands (fs, dfs, fsck, distcp etc)
> export HADOOP_CLIENT_OPTS="-Xmx512m $HADOOP_CLIENT_OPTS"
> {noformat}
> is merged into the variable HADOOP_OPTS in bin/hadoop, line 160:
> {noformat}
>     # Always respect HADOOP_OPTS and HADOOP_CLIENT_OPTS
>     HADOOP_OPTS="$HADOOP_OPTS $HADOOP_CLIENT_OPTS"
> {noformat}
> Orthogonally, the HADOOP_HEAPMAX environment variable triggers the setting of JAVA_HEAP_MAX in hadoop-config.sh, line 192:
> {noformat}
> # check envvars which might override default args
> if [ "$HADOOP_HEAPSIZE" != "" ]; then
>   #echo "run with heapsize $HADOOP_HEAPSIZE"
>   JAVA_HEAP_MAX="-Xmx""$HADOOP_HEAPSIZE""m"
>   #echo $JAVA_HEAP_MAX
> fi
> {noformat}
> Which is ultimately passed to the java application launcher in bin/hadoop, line 166:
> {noformat}
>     exec "$JAVA" $JAVA_HEAP_MAX $HADOOP_OPTS $CLASS "$@"
> {noformat}
> The result: two Xmx settings passed to the Java application launcher. Which one is respected depends on the JVM, but for HotSpot the rightmost setting is respected, and so the HADOOP_HEAPMAX environment variable has no effect and the default setting in HADOOP_CLIENT_OPTS wins. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)