add-attribute-index does not appear to launch its M/R job via yarn

add-attribute-index does not appear to launch its M/R job via yarn

The add-attribute-index command in the geomesa commandline tool does start the map-reduce job (AttributeIndexJob), but appears to run the job locally rather than submitting it via Yarn. This is the case despite the fact that the environment is properly configured to submit jobs via Yarn.

To add an attribute index for a test set of ~130M records on tcloud, this took about 8 hours. In contrast, submitting the job manually via the yarn commandline tool did submit the job to the cluster via yarn, and ultimately took less time (~2 hours) to complete.

Original JIRA Issue: https://geomesa.atlassian.net/browse/GEOMESA-1661

Key: GEOMESA-1661
Type: Bug
Priority: Major
Status: To Do
Resolution: Unresolved
Assignee: Matt Zimmerman
Reporter: Matt Zimmerman
Created: Thu, 2 Feb 2017 15:18:52 -0500
Updated: Thu, 2 Feb 2017 15:18:52 -0500

Start with the geomesa commandline tool's add-attribute-index command and trace how it launches AttributeIndexJob. Compare that path with the manual yarn command-line submission described in the issue. Done means the command submits the MapReduce job through YARN when the environment is configured for it, rather than running locally.