Hello,
If I understand your question correctly, you have cluster with 2 compute node, a head node and a broker node. Each compute node has 4 cores. You submit a WCF job with 4 tasks, and each task has min and max cores set to 4. Only 2 of
the tasks are running - the other 2 are queued.
If this is correct, then this behavior is expected. You cluster has only 2 compute nodes, and therefore only 8 cores on which tasks could run. Your job has 4 tasks, and each task requires 4 cores, for a total of 16 cores. Since only 8 cores
are available to run your tasks, the job scheduler cannot start all 4 tasks simultaneously. Instead, it starts only 2 tasks, and gives 4 cores to each of the tasks. The other 2 tasks are queued.
If you would like to be able to start all tasks in parallel, you can try to reduce the number of tasks in the job to 2, or reduce each task's resource requirement to 2 cores instead of 4, or, if possible, add more compute nodes to your cluster.
Best regards,
Leonid.
