Get Started With Ray on VMs

Install the cluster launcher, write your first cluster.yaml, and bring up a Ray cluster.

Install

pip install -U "ray[default]"

You’ll also need credentials for your cloud provider (aws configure, gcloud auth, az login).

Minimal cluster.yaml

cluster_name: hello-ray
provider:
  type: aws
  region: us-west-2
max_workers: 2
available_node_types:
  head:
    node_config:
      InstanceType: m5.large
  worker:
    min_workers: 1
    max_workers: 2
    node_config:
      InstanceType: m5.large
head_node_type: head
auth:
  ssh_user: ubuntu
setup_commands:
  - pip install "ray[default]==2.43.0"

Bring it up

ray up cluster.yaml

The launcher provisions VMs, installs Ray, and prints SSH commands.

Run a job

ray submit cluster.yaml -- python -c "import ray; ray.init(address='auto'); print(ray.cluster_resources())"

Or open a terminal on the head:

ray attach cluster.yaml

Forward the dashboard

ray dashboard cluster.yaml

Forwards port 8265 from the head to your local machine.

Tear it down

ray down cluster.yaml

Next steps

AWS

AWS-specific options.

Autoscaling

Configure scaling behavior.