Step 5: Validate Cluster Configuration and Status

The Drill-on-YARN command line tool prints a URL, for the Drill Application Master process, that you can use to monitor the cluster. The Drillbits should be up and running, unless the upload or launch failed.

Failed upload: If the upload fails, the most likely reason is that the HADOOP_HOME variable is not set, or the user launching Drill-on-YARN does not have permission to create or write to the DFS location set in the Drill-on-YARN configuration file.
Failed launch: If the launch fails, verify that the YARN maximum container size was set to be at least as large as the Drill container size and that YARN can provide the number of vcores and disks you have requested for Drill.

Monitor Cluster Activity

When the cluster launches successfully, you can verify the status of the components in the Drill cluster from the command line or you can copy the Application Master URL into your browser to use the Application Master web UI.

To check the status from the command line, issue the following command:

$DRILL_HOME/bin/drill-on-yarn.sh --site $DRILL_SITE status

Verify that the Drillbits are in the Running state (or transition to that state after a few moments.) If the Drillbits remain in the Requesting state, the likely cause is that YARN cannot provide a container of the requested size. You can access this information on the Drillbits page in the Application Master web UI.
Verify that the Application Master has correctly picked up the YARN-related configuration from drill-override.conf and drill-on-yarn.conf. You can access this information on the Configuration page in the Application Master web UI.
Verify that the Application Master is running. The Drill Application Master uses ZooKeeper to verify that only one Application Master runs per Drill cluster. The Application Master fails if ZooKeeper is down, is configured incorrectly in drill-override.conf, or if another Application Master is running for the same cluster.
Verify that that the configurations in are correct in drill-override.conf if there are Drillbit failures. Drill-on-YARN detects Drillbit failures and retries each Drillbit a few times. If the Drillbit continues to fail, the node is black-listed for that run of Drill-on-YARN. You can see failures in the History page of the web UI. If your Drillbits fail, the most likely reason is misconfiguration within drill-override.conf. Use YARN to locate and view the Drill logs for the failed container.

Resize the Cluster

If the cluster works well for a single node, you can increase the number of nodes.

NOTE In the Drill 1.8 release, adding nodes can be done while users run queries. However, removing nodes from the cluster causes queries to fail, and so should only be done when the cluster is idle.

To make a permanent change, modify the cluster size in drill-on-yarn.conf:

cluster: [
   {
     ...
     count: 5
   }
 ]

You can also resize the cluster dynamically. To add two nodes from the command line:

$DRILL_HOME/bin/drill-on-yarn.sh --site $DRILL_SITE resize +2

To set the cluster size to a total of five nodes:

$DRILL_HOME/bin/drill-on-yarn.sh --site $DRILL_SITE resize 5

To remove one of the nodes:

$DRILL_HOME/bin/drill-on-yarn.sh --site $DRILL_SITE resize -1

You can also resize your cluster using the Manage page of the Application Master web UI, by entering the number of desired nodes and clicking Go.

Stop the Cluster

You can stop the cluster from the command line or from the Manage page in the Application Master web UI.

To stop the cluster from the command line, issue the following command:

$DRILL_HOME/bin/drill-on-yarn.sh --site $DRILL_SITE stop