ray/python/ray/autoscaler/staroid/example-full.yaml

# An unique identifier for the head node and workers of this cluster.
# A namespace will be automatically created for each cluster_name in SKE.
cluster_name: default # name with 'a-z' and '-'

# The minimum number of workers nodes to launch in addition to the head
# node. This number should be >= 0.
min_workers: 0

# The maximum number of workers nodes to launch in addition to the head
# node. This takes precedence over min_workers.
max_workers: 5

# The initial number of worker nodes to launch in addition to the head
# node. When the cluster is first brought up (or when it is refreshed with a
# subsequent `ray up`) this number of nodes will be started.
initial_workers: 0

# Whether or not to autoscale aggressively. If this is enabled, if at any point
#   we would start more workers, we start at least enough to bring us to
#   initial_workers.
autoscaling_mode: default

# The autoscaler will scale up the cluster to this target fraction of resource
# usage. For example, if a cluster of 10 nodes is 100% busy and
# target_utilization is 0.8, it would resize the cluster to 13. This fraction
# can be decreased to increase the aggressiveness of upscaling.
# This value must be less than 1.0 for scaling to happen.
target_utilization_fraction: 0.8

# If a node is idle for this many minutes, it will be removed.
idle_timeout_minutes: 5

# Kubernetes resources that need to be configured for the autoscaler to be
# able to manage the Ray cluster. If any of the provided resources don't
# exist, the autoscaler will attempt to create them. If this fails, you may
# not have the required permissions and will have to request them to be
# created by your cluster administrator.
provider:
    type: staroid

    # Access token for Staroid from https://staroid.com/settings/accesstokens.
    # Alternatively, you can set STAROID_ACCESS_TOKEN environment variable.
    # https://github.com/staroids/staroid-python#configuration
    # for more information.
    access_token:

    # Staroid account to use. e.g. GITHUB/staroids
    # Alternatively, you can set STAROID_ACCOUNT environment variable.
    # Leave empty to select default account for given access token.
    # https://github.com/staroids/staroid-python#configuration
    # for more information.
    account:

    # Name of a Staroid Kubernetes Engine (SKE) instance.
    # Alternatively, you can set STAROID_SKE environment variable.
    # An SKE is a virtualized Kubernetes cluster.
    # Will create a new if not exists.
    ske: "Ray cluster"

    # Cloud and Region to create an SKE when not exists.
    # If SKE already exists, this value will be ignored.
    # Supported cloud region can be found
    # https://docs.staroid.com/ske/cloudregion.html.
    ske_region: "aws us-west2"

    # To create a namespace in SKE, you need to specify a Github project.
    # The Github project needs to have a staroid.yaml
    # (https://docs.staroid.com/references/staroid_yaml.html).
    # staroid.yaml defines various resources for the project, such as
    #   - Building container images can be accessed from the namespace
    #   - Kubernetes resources to create (like Persistent volume claim)
    #     on namespace creation
    # You can fork when you need to customize.
    #   1. Fork github.com/open-datastudio/ray-cluster
    #   2. Change contents
    #   3. Connect forked repository (https://staroid.com/projects/settings)
    #   4. Release your customized branch
    #      4-1. Select project from 'My projects' menu
    #      4-2. Select your branch in 'Release' tab
    #      4-3. After build success, switch to 'Production'
    #      4-4. Switch Launch permission to 'Public' if required
    #   5. Change 'project' field to point your 
    #      repository and branch in this file
    project: "GITHUB/open-datastudio/ray-cluster:master"

    # 'spec.containers.image' field for ray-node and ray-worker will be
    # overrided by the image built from the 'project' field above.
    # Set this value to 'false' to not override the image.
    image_from_project: true

    # Python version to use. One of '3.6.9', '3.7.7', '3.8.3'.
    # 'project' field above provides docker image for each python version.
    # Fork 'project' if you'd like to support other python versions.
    python_version: 3.7.7

    # Exposing external IP addresses for ray pods isn't currently supported.
    use_internal_ips: true

# Kubernetes pod config for the head node pod.
head_node:
    apiVersion: v1
    kind: Pod
    metadata:
        # Automatically generates a name for the pod with this prefix.
        generateName: ray-head-

        # Must match the head node service selector above if a head node
        # service is required.
        labels:
            component: ray-head

            # https://docs.staroid.com/ske/pod.html
            pod.staroid.com/spot: "false" # use on-demand instance for head.

            # Locate ray head to dedicated Kubernetes node
            # In dedicated mode, resource requests and limits in the pod spec will be
            # automatically overrided based on 'pod.staroid.com/instance-type' below.
            pod.staroid.com/isolation: dedicated # 'sandboxed' or 'dedicated'

            # Instance type to use in 'dedicated' mode, such as 'standard-4', 'gpu-1'.
            # See available instance type from https://docs.staroid.com/ske/pod.html.
            pod.staroid.com/instance-type: standard-4
    spec:
        automountServiceAccountToken: true

        # Restarting the head node automatically is not currently supported.
        # If the head node goes down, `ray up` must be run again.
        restartPolicy: Never

        # This volume allocates shared memory for Ray to use for its plasma
        # object store. If you do not provide this, Ray will fall back to
        # /tmp which cause slowdowns if is not a shared memory volume.
        volumes:
        - name: dshm
          emptyDir:
              medium: Memory
        - name: tmp-volume
          emptyDir: {}
        # nfs volume provides a shared volume across all ray-nodes.
        - name: nfs-volume
          persistentVolumeClaim:
              claimName: nfs

        containers:
        - name: ray-node
          imagePullPolicy: Always
          # You are free (and encouraged) to use your own container image,
          # but it should have the following installed:
          #   - rsync (used for `ray rsync` commands and file mounts)
          #   - screen (used for `ray attach`)
          #   - kubectl (used by the autoscaler to manage worker pods)
          # Image will be overridden when 'image_from_project' is true.
          image: rayproject/autoscaler
          # Do not change this command - it keeps the pod alive until it is
          # explicitly killed.
          command: ["/bin/bash", "-c", "--"]
          args: ["touch ~/.bashrc; trap : TERM INT; sleep infinity & wait;"]
          ports:
              - containerPort: 6379 # Redis port.
              - containerPort: 6380 # Redis port.
              - containerPort: 6381 # Redis port.
              - containerPort: 12345 # Ray internal communication.
              - containerPort: 12346 # Ray internal communication.

          # This volume allocates shared memory for Ray to use for its plasma
          # object store. If you do not provide this, Ray will fall back to
          # /tmp which cause slowdowns if is not a shared memory volume.
          volumeMounts:
              - mountPath: /dev/shm
                name: dshm
              - mountPath: /tmp
                name: tmp-volume
              - mountPath: /nfs
                name: nfs-volume
          resources:
              requests:
                  cpu: 4000m
                  memory: 8Gi
              limits:
                  cpu: 4000m
                  # The maximum memory that this pod is allowed to use. The
                  # limit will be detected by ray and split to use 10% for
                  # redis, 30% for the shared memory object store, and the
                  # rest for application memory. If this limit is not set and
                  # the object store size is not set manually, ray will
                  # allocate a very large object store in each pod that may
                  # cause problems for other pods.
                  memory: 8Gi
          env:
              # This is used in the head_start_ray_commands below so that
              # Ray can spawn the correct number of processes. Omitting this
              # may lead to degraded performance.
              - name: MY_CPU_REQUEST
                valueFrom:
                    resourceFieldRef:
                        resource: limits.cpu
              - name: RAY_ADDRESS
                value: "auto"

# Kubernetes pod config for worker node pods.
worker_nodes:
    apiVersion: v1
    kind: Pod
    metadata:
        # Automatically generates a name for the pod with this prefix.
        generateName: ray-worker-

        # Must match the worker node service selector above if a worker node
        # service is required.
        labels:
            component: ray-worker

            # https://docs.staroid.com/ske/pod.html
            pod.staroid.com/spot: "true"

            # Locate ray head to dedicated Kubernetes node
            # In dedicated mode, resource requests and limits in the pod spec will be
            # automatically overrided based on 'pod.staroid.com/instance-type' below.
            pod.staroid.com/isolation: dedicated # 'sandboxed' or 'dedicated'

            # Instance type to use in 'dedicated' mode, such as 'standard-4', 'gpu-1'.
            # See available instance type from https://docs.staroid.com/ske/pod.html.
            pod.staroid.com/instance-type: standard-4
    spec:
        serviceAccountName: default

        # Worker nodes will be managed automatically by the head node, so
        # do not change the restart policy.
        restartPolicy: Never

        # This volume allocates shared memory for Ray to use for its plasma
        # object store. If you do not provide this, Ray will fall back to
        # /tmp which cause slowdowns if is not a shared memory volume.
        volumes:
        - name: dshm
          emptyDir:
              medium: Memory
        - name: tmp-volume
          emptyDir: {}
        - name: nfs-volume
          persistentVolumeClaim:
              claimName: nfs
        containers:
        - name: ray-node
          imagePullPolicy: Always
          # You are free (and encouraged) to use your own container image,
          # but it should have the following installed:
          #   - rsync (used for `ray rsync` commands and file mounts)
          image: rayproject/autoscaler
          # Do not change this command - it keeps the pod alive until it is
          # explicitly killed.
          command: ["/bin/bash", "-c", "--"]
          args: ["touch ~/.bashrc; trap : TERM INT; sleep infinity & wait;"]
          ports:
              - containerPort: 12345 # Ray internal communication.
              - containerPort: 12346 # Ray internal communication.

          # This volume allocates shared memory for Ray to use for its plasma
          # object store. If you do not provide this, Ray will fall back to
          # /tmp which cause slowdowns if is not a shared memory volume.
          volumeMounts:
              - mountPath: /dev/shm
                name: dshm
              - mountPath: /tmp
                name: tmp-volume
              - mountPath: /nfs
                name: nfs-volume
          resources:
              requests:
                  cpu: 4000m
                  memory: 8Gi
              limits:
                  cpu: 4000m
                  # This memory limit will be detected by ray and split into
                  # 30% for plasma, and 70% for workers.
                  memory: 8Gi
          env:
              # This is used in the head_start_ray_commands below so that
              # Ray can spawn the correct number of processes. Omitting this
              # may lead to degraded performance.
              - name: MY_CPU_REQUEST
                valueFrom:
                    resourceFieldRef:
                        resource: limits.cpu

# Files or directories to copy to the head and worker nodes. The format is a
# dictionary from REMOTE_PATH: LOCAL_PATH, e.g.
file_mounts: {
#    "/path1/on/remote/machine": "/path1/on/local/machine",
#    "/path2/on/remote/machine": "/path2/on/local/machine",
}

# Files or directories to copy from the head node to the worker nodes. The format is a
# list of paths. The same path on the head node will be copied to the worker node.
# This behavior is a subset of the file_mounts behavior. In the vast majority of cases
# you should just use file_mounts. Only use this if you know what you're doing!
cluster_synced_files: []

# List of commands that will be run before `setup_commands`. If docker is
# enabled, these commands will run outside the container and before docker
# is setup.
initialization_commands: []

# List of shell commands to run to set up nodes.
setup_commands: []

# Custom commands that will be run on the head node after common setup.
head_setup_commands:
    # install staroid and kubernetes packages. Staroid node provider depends on them which autoscaler will use.
    - pip install -q staroid kubernetes
    # install jupyterlab
    - pip install -q jupyterlab
    - ln -s /nfs /home/ray/nfs
    - bash -c 'jupyter-lab --ip="*" --NotebookApp.token="" --NotebookApp.password="" --NotebookApp.allow_origin="*" --NotebookApp.notebook_dir="/home/ray"' &
    # show 'notebook' link in staroid management console to access jupyter notebook.
    - 'echo -e "kind: Service\napiVersion: v1\nmetadata:\n  name: notebook\n  annotations:\n    service.staroid.com/link: show\nspec:\n  ports:\n  - name: http\n    port: 8888\n  selector:\n    component: ray-head" | kubectl apply -f -'

# Custom commands that will be run on worker nodes after common setup.
worker_setup_commands: []

# Command to start ray on the head node. You don't need to change this.
# Note webui-host is set to 0.0.0.0 so that kubernetes can port forward.
head_start_ray_commands:
    - ray stop
    - ulimit -n 65536; ray start --head --num-cpus=$MY_CPU_REQUEST --port=6379 --object-manager-port=8076 --autoscaling-config=~/ray_bootstrap_config.yaml --dashboard-host 0.0.0.0

# Command to start ray on worker nodes. You don't need to change this.
worker_start_ray_commands:
    - ray stop
    - ulimit -n 65536; ray start --num-cpus=$MY_CPU_REQUEST --address=$RAY_HEAD_IP:6379 --object-manager-port=8076