Name compute resources with the amount of memory and number of cores (#…

…265)

Instead of naming the queue and the CR the same, after the instance type,
name the CR with the instance type memory and core amounts to give cluster
users more information about the compute nodes.
Many users will not know how much memory and how many cores are available
based on the instance type name.

Resolves #264

source/cdk/cdk_slurm_stack.py

@@ -2497,12 +2497,20 @@ def create_parallel_cluster_config(self):
         number_of_queues = 0
         number_of_compute_resources = 0
 
+        disable_smt = self.config['slurm']['ParallelClusterConfig']['DisableSimultaneousMultithreading']
+
         # Create 1 queue and compute resource for each instance type and purchase option.
+        # The queue is named after the instance type.
+        # The CR is named after the amount of memory and number of cores.
         for purchase_option in purchase_options:
             for instance_type in self.instance_types:
                 logger.debug(f"Creating queue for {purchase_option} {instance_type}")
                 efa_supported = self.plugin.get_EfaSupported(self.cluster_region, instance_type) and self.config['slurm']['ParallelClusterConfig']['EnableEfa']
                 mem_gb = int(self.plugin.get_MemoryInMiB(self.cluster_region, instance_type) / 1024)
+                core_count = int(self.plugin.get_CoreCount(self.cluster_region, instance_type))
+                threads_per_core = int(self.plugin.get_DefaultThreadsPerCore(self.cluster_region, instance_type))
+                if not disable_smt:
+                    core_count *= threads_per_core
                 if purchase_option == 'ONDEMAND':
                     queue_name_prefix = "od"
                     allocation_strategy = 'lowest-price'
@@ -2524,6 +2532,7 @@ def create_parallel_cluster_config(self):
                 if number_of_queues >= MAX_NUMBER_OF_QUEUES:
                     logger.error(f"Can't create {queue_name} queue because MAX_NUMBER_OF_QUEUES=={MAX_NUMBER_OF_QUEUES} and have {number_of_queues} queues.")
                     exit(1)
+                # ParallelCluster creates a NodeSet for each queue that contains all NodeNames in the queue.
                 nodeset = f"{queue_name}_nodes"
                 if purchase_option_partition not in partition_nodesets:
                     partition_nodesets[purchase_option_partition] = []
@@ -2532,12 +2541,20 @@ def create_parallel_cluster_config(self):
                 if mem_partition not in partition_nodesets:
                     partition_nodesets[mem_partition] = []
                 partition_nodesets[mem_partition].append(nodeset)
+                mem_core_partition = f"{queue_name_prefix}-{mem_gb}-gb-{core_count}-cores"
+                if mem_core_partition not in partition_nodesets:
+                    partition_nodesets[mem_core_partition] = []
+                partition_nodesets[mem_core_partition].append(nodeset)
                 parallel_cluster_queue = self.create_queue_config(queue_name, allocation_strategy, purchase_option)
                 number_of_queues += 1
 
-                compute_resource_name = f"{queue_name_prefix}-{instance_type}".replace('.', '-')
-                compute_resource_name = compute_resource_name.replace('large', 'l')
-                compute_resource_name = compute_resource_name.replace('medium', 'm')
+                if True:
+                    # CR must begin with an alpha character, otherwise don't need the queue_name_prefix
+                    compute_resource_name = f"{queue_name_prefix}-{mem_gb}-gb-{core_count}-cores"
+                else:
+                    compute_resource_name = f"{queue_name_prefix}-{instance_type}".replace('.', '-')
+                    compute_resource_name = compute_resource_name.replace('large', 'l')
+                    compute_resource_name = compute_resource_name.replace('medium', 'm')
                 if number_of_compute_resources >= MAX_NUMBER_OF_COMPUTE_RESOURCES:
                     logger.error(f"Can't create {compute_resource_name} compute resource because MAX_NUMBER_OF_COMPUTE_RESOURCES=={MAX_NUMBER_OF_COMPUTE_RESOURCES} and have {number_of_compute_resources} compute resources")
                     exit(1)