[
https://issues.apache.org/jira/browse/HADOOP-13935?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15771987#comment-15771987
]
Allen Wittenauer commented on HADOOP-13935:
-------------------------------------------
It's a pretty common configuration to not define JAVA_HOME in hadoop-env.sh but
as part of a /etc/profile.d configuration. In the future, we're likely to see
.hadoop-env and .hadooprc set it. So it isn't safe to make the assumption that
this is being defined there.
Experience with Apache Yetus has shown that the binary location is actually
much more important than the actual version being reported when debugging these
types of issues. A machine with different JDKs (e.g., IBM, OpenJDK, Oracle,
etc) may report the same version string but will always have different paths.
The path (resolved or otherwise) is also likely to have the version string as
part of the directory component.
That said, this issue is basically covered by HADOOP-12366. As part of that
JIRA issue, the various envvars subcommands in hadoop, hdfs, mapred, and yarn
expose what JAVA_HOME is in use when the shell scripts launch Java commands
(barring a custom JAVA_HOME getting set via dynamic subcommands).
> A command to print JAVA_VERSION used by Hadoop/HDFS
> ---------------------------------------------------
>
> Key: HADOOP-13935
> URL: https://issues.apache.org/jira/browse/HADOOP-13935
> Project: Hadoop Common
> Issue Type: Improvement
> Components: common
> Affects Versions: 3.0.0-alpha2
> Reporter: Anu Engineer
> Assignee: Yiqun Lin
> Labels: newbie
>
> The java version used by hadoop is controlled by JAVA_HOME variable defined
> in hadoop_env.sh. We log this information when HDFS starts in the log file.
> {noformat}
> STARTUP_MSG: java = 1.8.0_112
> {noformat}
> However, it is quite possible that a user might have many versions of java
> installed on his/her machine. Generally, users tend to check for the java
> version via
> {noformat}
> java -version
> {noformat}
> This just means we are printing out the java version in the current shell
> path.
> This jira proposes adding a new simple command or an extension to existing
> hadoop version command where the current java version used by hadoop is also
> printed out.
> This avoids customer confusion when they are looking at if the java stack is
> properly configured. For example, checking if JCE is installed correctly.
> This is a very minor change that can be done by modifying hdfs.cmd or hdfs
> shell script in the /bin directory. Thanks to [~sujit] for bringing this to
> my attention.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
