ray-aws-cluster.yaml

# An unique identifier for the head node and workers of this cluster.
cluster_name: rayz1d

# Cloud-provider specific configuration.
provider:
    type: aws
    region: us-east-1
    cache_stopped_nodes: True

max_workers: 30

setup_commands:
    - sudo bash -c 'rm -rf /etc/security/limits.d; echo "* soft nofile 65535" >> /etc/security/limits.conf; echo "* hard nofile 65535" >> /etc/security/limits.conf;'
    - sudo bash -c 'mkswap /dev/nvme1n1 && swapon /dev/nvme1n1'
    # - sudo apt install swig -y
    # - pip install -U 'boto3>=1.4.8'
    # - pip install -U 'ray[default,rllib]==2.0.0'
    # - pip install -U tensorflow==2.10.0 tensorflow-probability==0.17.0 torch==1.12.1 wandb 'pettingzoo[magent,sisl]==1.20.1' supersuit==3.5.0
    # - pip install -U gym==0.22.0


available_node_types:
    ray.head.default:
        resources: {"CPU": 1, "object-store-memory": 0}
        node_config:
            InstanceType: z1d.12xlarge
            ImageId: ami-0c3a82216886b746f # My AMI with prerequisites installed
    ray.worker.default:
        min_workers: 0
        max_workers: 16
        resources: {"object-store-memory": 0}
        node_config:
            InstanceType: z1d.6xlarge
            ImageId: ami-0c3a82216886b746f # My AMI with prerequisites installed
    ray.worker.3xlarge:
        min_workers: 0
        max_workers: 16
        resources: {"object-store-memory": 0}
        node_config:
            InstanceType: z1d.3xlarge
            ImageId: ami-0c3a82216886b746f # My AMI with prerequisites installed
    ray.worker.2xlarge:
        min_workers: 0
        max_workers: 16
        resources: {"object-store-memory": 0}
        node_config:
            InstanceType: z1d.2xlarge
            ImageId: ami-0c3a82216886b746f # My AMI with prerequisites installed
    ray.worker.12xlarge:
        min_workers: 0
        max_workers: 16
        resources: {"object-store-memory": 0}
        node_config:
            InstanceType: z1d.12xlarge
            ImageId: ami-0c3a82216886b746f # My AMI with prerequisites installed

# Specify the node type of the head node (as configured above).
head_node_type: ray.head.default