Continuous scheduling is known to cause locking issue inside
the scheduler in larger cluster, more than 100 nodes, use
FairSchedulerConfiguration.ASSIGN_MULTIPLE to improve container allocation ramp up.
The preferred way to configure the increment is by using the
yarn.resource-types.{RESOURCE_NAME}.increment-allocation property,
for memory: yarn.resource-types.memory-mb.increment-allocation
The preferred way to configure the increment is by using the
yarn.resource-types.{RESOURCE_NAME}.increment-allocation property,
for CPU: yarn.resource-types.vcores.increment-allocation
Continuous scheduling should not be turned ON. It is
deprecated because it can cause scheduler slowness due to locking issues.
Schedulers should use assignmultiple as a replacement.