maxConcurrencyPerNode vs Cluster.getWorkerMaxConcurrentComputes

How does the priority work between setting maxConcurrencyPerNode on a BatchJob/MapReduce job vs the Cluster-level setting for max concurrent computes? Is it always the lower number of the two?

Specifically thinking of two different scenarios:

maxConcurrencyPerNode < Cluster.getWorkerMaxConcurrentComputes()

I would think maxConcurrencyPerNode wins out here

maxConcurrencyPerNode > Cluster.getWorkerMaxConcurrentComputes()

I would think Cluster.getWorkerMaxConcurrentComputes wins out here

1 Like

@scott.kruyswyk The cluster setting limits the number of threads that can do any invalidation queue computing. The job setting then limits things further but does not allow you to exceed the cluster setting.

1 Like