Step 4: Launch Drill Under YARN
Now that the Drill and YARN configuration is complete, you can issue the start
command from the Drill-on-YARN client to launch Drill under YARN. Launching Drill-on-YARN from
the client starts Drill and brings Drill up on other nodes.
Issuing the start command starts the YARN Application Master, which then works with YARN to
start the Drillbits. The Application Master provides a web UI to monitor the cluster.
NOTE
To
simplify debugging, you can set the cluster size to a single node. Once you confirm that a
single node works, increase the node count.Launch Drill under YARN as the
mapr
user. For example, if you installed
Drill as mapr
launch Drill as the mapr
user. NOTE
If you
launch Drill as root
and the system returns an error failing the launch
attempt, launch Drill as a user with permissions, such as mapr
.Issue the following command to start Drill under
YARN:
$DRILL_HOME/bin/drill-on-yarn.sh --site $DRILL_SITE start
NOTE
To
run SQLLine, you must also add the --site
option:$DRILL_HOME/bin/sqlline --site $DRILL_SITE
TIP
To avoid typing the site
argument each time you launch Drill
under YARN, set an environment
variable:export DRILL_CONF_DIR=$DRILL_SITE
$DRILL_HOME/bin/drill-on-yarn.sh start
After you issue the start command, a number of lines describing the start-up process print. The
tool automatically archives and uploads the site directory, which YARN copies (along with the
Drill software) onto each node. A URL that includes both the host and the port number
displays. Enter the URL in a web browser to access the Application Master web UI.
NOTE
When
you launch Drill from the Drill-on-YARN client, the Application Master can come up on any
node. Save the provided URL to share with other users so they can also access the
Application Master. Alternatively, you can run the status
command to see
the URL or go to the YARN Resource Manager UI to get the link.See Drill-on-YARN Command Line Tool for additional commands, including stop, status, and resize. See Application Master Web UI for cluster monitoring information.