You can configure the Blaze engine to submit Blaze jobs to a YARN queue that is configured with a set of node labels. You can simultaneously configure the Blaze engine using a node label to start the engine on the cluster nodes that have the node label. The node label that you configure for the Blaze engine must be accessible by the YARN queue that receives submitted Blaze jobs.

For example, the Hadoop cluster might be divided into the following partitions:

memory. Partition of nodes that have higher memory.
fast. Partition of nodes that have higher CPU.
default. Default partition.

You configure a YARN queue

Blaze_only

using the node labels

fast

and

default

. When you configure the Blaze engine, you specify the queue

Blaze_only

and the node label

fast

. Note that the YARN queue can access the node label

fast

.

The following image shows the configurations for the Blaze engine in the Hadoop connection:

The following image shows the nodes where the Blaze engine starts and the nodes that are accessible to the queue:

The Blaze engine starts on the nodes in the
fast
partition.
The queue can access the
fast
and the
default
partitions. The jobs in the queue run on nodes in the
fast
and the
default
partitions.