[
https://issues.apache.org/jira/browse/MAHOUT-426?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Drew Farris updated MAHOUT-426:
-------------------------------
Attachment: MAHOUT-426.patch
Thanks for the patch Peter. All very good points.
Attached is a slight revision to Peter's patch that does the following:
* Adds the existence check for MAHOUT_JOB when looking in
$MAHOUT_HOME/mahout-examples-*.job as Peter originally proposed.
* Adds documentation for the -core option in the header
* Adds MAHOUT_CORE environment as an alternative to the -core option
* Instead of automatically falling back to local mode If MAHOUT_JOB is not set
because the mahout-examples-*.job file can't be found when HADOOP_HOME and
HADOOP_CONF_DIR are set, the script will error out indicating that the examples
job can't be found with a helpful suggestion that 'mvn install' should be run.
Additionally, added the MAHOUT_LOCAL environment variable to force mahout to
run in local mode although HADOOP_HOME and HADOOP_CONF_DIR are set. This is
useful for developers to set in conjunction with MAHOUT_CORE
I'll pause a bit to give the chance for people to review and then commit.
> The Mahout command script doesn't properly run when Hadoop is installed but
> $MAHOUT_JOB is not set or found in standard locations
> ---------------------------------------------------------------------------------------------------------------------------------
>
> Key: MAHOUT-426
> URL: https://issues.apache.org/jira/browse/MAHOUT-426
> Project: Mahout
> Issue Type: Bug
> Affects Versions: 0.4
> Environment: Manifested in Amazon EC2 instance set up according to
> wiki instructions - https://cwiki.apache.org/MAHOUT/mahoutec2.html
> Seems likely to affect any server where Hadoop is installed
> Reporter: Peter Goldstein
> Attachments: MAHOUT-426.patch, MAHOUT-426.patch
>
>
> In the case where no $MAHOUT_JOB is set, and no mahout-examples-*.job file is
> found in the standard locations, the Mahout command script should default to
> the run locally option. Instead, when the script is run one received an
> error that the non-existent file "$MAHOUT_HOME/mahout-examples-*.job" cannot
> be found.
> This is the result of two separate bugs in the Mahout command script:
> i) There is no existence check in the loop checking for mahout-examples-*.job
> files in $MAHOUT_HOME
> ii) The execution if/else clause only checks if either $HADOOP_CONF_DIR or
> $HADOOP_HOME is empty when determining whether to run locally. It also
> should check the $MAHOUT_JOB variable, as it is meaningless to run hadoop in
> this case.
> Fixing these two issues allowed me to (almost) run the
> $MAHOUT_HOME/examples/bin/build-reuters.sh script.
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
