Launch instances with On-Demand Capacity Reservations (ODCR)

Create a resource group, to group capacity.

$ aws resource-groups create-group --name EC2CRGroup \ 
    --configuration '{"Type":"AWS::EC2::CapacityReservationPool"}' '{"Type":"AWS::ResourceGroups::Generic", "Parameters": [{"Name": "allowed-resource-types", "Values": ["AWS::EC2::CapacityReservation"]}]}'

#!/bin/bash
set -e

# Override run_instance attributes
cat > /opt/slurm/etc/pcluster/run_instances_overrides.json << EOF
{
    "my-queue": {
        "my-compute-resource": {
            "CapacityReservationSpecification": {
                "CapacityReservationTarget": {
                    "CapacityReservationId": "cr-abcdef01234567890"
                }
            }
        }
    }
}
EOF

Add capacity reservations to the resource group. Every time that you create a new ODCR, add it to the Group Reservation. Replace ACCOUNT_ID with your account ID, PLACEHOLDER_CAPACITY_RESERVATION with your capacity reservation ID, and REGION_ID with your AWS Region ID (for example, us-east-1).

$ aws resource-groups group-resources --region REGION_ID --group EC2CRGroup \
    --resource-arns arn:aws:ec2:REGION_ID:ACCOUNT_ID:capacity-reservation/PLACEHOLDER_CAPACITY_RESERVATION

Create a policy document on your local computer. Replace ACCOUNT_ID with your account ID and REGION_ID with your AWS Region ID (for example, us-east-1).

cat > policy.json << EOF
{
    "Version": "2012-10-17",		 	 	 
    "Statement": [
        {
            "Sid": "RunInstancesInCapacityReservation",
            "Effect": "Allow",
            "Action": "ec2:RunInstances",
            "Resource": [
                "arn:aws:ec2:REGION_ID:ACCOUNT_ID:capacity-reservation/*",
                "arn:aws:resource-groups:REGION_ID:ACCOUNT_ID:group/*"
            ]
        }
    ]
}
EOF

Create the IAM policy on your AWS account using the json file that you created.

$ aws iam create-policy --policy-name RunInstancesCapacityReservation --policy-document file://policy.json

Create the following post install script locally on the instance and name it postinstall.sh.

Replace ACCOUNT_ID with your AWS account ID, and REGION_ID with your AWS Region ID (for example, us-east-1).

#!/bin/bash
set -e

# Override run_instance attributes
cat > /opt/slurm/etc/pcluster/run_instances_overrides.json << EOF
{
    "my-queue": {
        "my-compute-resource": {
            "CapacityReservationSpecification": {
                "CapacityReservationTarget": {
                    "CapacityReservationResourceGroupArn": "arn:aws:resource-groups:REGION_ID:ACCOUNT_ID:group/EC2CRGroup"
                }
            }
        }
    }
}
EOF

Upload the file to an Amazon S3 bucket. Replace amzn-s3-demo-bucket with your specific S3 bucket name.

$ aws s3 mb s3://amzn-s3-demo-bucket
aws s3 cp postinstall.sh s3://amzn-s3-demo-bucket/postinstall.sh

Create the local cluster configuration, replacing the placeholders with your own values.

Region: REGION_ID
Image:
  Os: alinux2
HeadNode:
  InstanceType: c5.2xlarge
  Ssh:
    KeyName: YOUR_SSH_KEY
  Iam:
    S3Access:
      - BucketName: amzn-s3-demo-bucket
    AdditionalIamPolicies:
      - Policy: arn:aws:iam::ACCOUNT_ID:policy/RunInstancesCapacityReservation
  ## This post-install script is executed after the node is configured.
  ## It is used to install scripts at boot time and specific configurations
  ## In the script below we are overriding the calls to RunInstance to force
  ## the provisioning of our my-queue partition to go through
  ## the On-Demand Capacity Reservation
  CustomActions:
    OnNodeConfigured:
      Script: s3://amzn-s3-demo-bucket/postinstall.sh
  Networking:
    SubnetId: YOUR_PUBLIC_SUBNET_IN_TARGET_AZ

Scheduling:
  Scheduler: slurm
  SlurmQueues:
    - Name: my-queue
      ComputeResources:
        - MinCount: 0
          MaxCount: 100
          InstanceType: p4d.24xlarge
          Name: my-compute-resource
          Efa:
            Enabled: true
      Networking:
        ## PlacementGroup:
        ##   Enabled: true ## Keep PG disabled if using targeted ODCR
        SubnetIds:
          - YOUR_PRIVATE_SUBNET_IN_TARGET_AZ

Create the cluster.

Use the following command to create the cluster. Replace cluster-config.yaml with your configuration file name, cluster-dl with your cluster name, and REGION_ID with your Region ID (for example, us-east-1).

$ pcluster create-cluster --cluster-configuration cluster-config.yaml --cluster-name cluster-dl --region REGION_ID

After the cluster is created, the post-install script runs in the head node. The script creates the run_instances_overrides.json file and overrides the calls to RunInstances to force the provisioning of the partition to go through the On-Demand Capacity Reservation.

The AWS ParallelCluster daemons that are responsible for cluster scaling automatically use this configuration for new instances that are launched. To confirm that the specified parameters are being used to provision instances, you can look at the following log files:

If the parameters are correct, you'll find a log entry contains the following.

Found RunInstances parameters override. Launching instances with: <parameters_list>