ray/deploy/ray-operator/config/samples/ray_v1_raycluster.complete.yaml

apiVersion: ray.io/v1alpha1
kind: RayCluster
metadata:
  labels:
    controller-tools.k8s.io: "1.0"
    # An unique identifier for the head node and workers of this cluster.
  name: raycluster-sample
spec:
  # An unique identifier for the head node and workers of this cluster.
  clusterName: raycluster-sample
  images:
    defaultImage: rayproject/autoscaler
  imagePullPolicy: "Always"

  extensions:
    # the pod replicas in this group typed worker
    - replicas: 1
      # logical group name, for this called small-group, also can be functional
      groupName: small-group
      # pod type
      type: worker

      # Command to start ray
      command: ray start --block --node-ip-address=$MY_POD_IP --address=$CLUSTER_NAME-head:6379 --object-manager-port=12345 --node-manager-port=12346 --object-store-memory=100000000 --num-cpus=1

      # custom labels. NOTE: do not define custom labels start with `raycluster.`, they may be used in controller.
      # Refer to https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
      labels:
        raycluster.group.name: small-group

      # annotations for pod
      annotations:
        key: value

      # use affinity to select nodes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: key1
                    operator: In
                    values: ["true"]
                  - key: key2
                    operator: In
                    values: ["true"]

      # use tolerations to select nodes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
      tolerations:
        - key: "key1"
          operator: "Equal"
          value: "true"
          effect: "NoSchedule"

      # resource requirements.
      # Refer to https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
      resources:
        limits:
          cpu: 1000m
          memory: 2Gi
          ephemeral-storage: 2Gi
        requests:
          cpu: 1000m
          memory: 2Gi
          ephemeral-storage: 2Gi

      # environment variables to set in the container.Optional.
      # Refer to https://kubernetes.io/docs/tasks/inject-data-application/define-environment-variable-container/
      containerEnv:
        - name: APPNAME
          value: raycluster-sample
        - name: MY_POD_NAME
          valueFrom:
            fieldRef:
              fieldPath: metadata.name
        - name: MY_POD_IP
          valueFrom:
            fieldRef:
              fieldPath: status.podIP

      # head service suffix: {namespace}.svc , follows Kubernetes standard
      headServiceSuffix: "ray-operator.svc"

      # use volumes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/storage/volumes/
      volumes:
        - name: log-volume
          emptyDir: {}

      # use volumeMounts.Optional.
      # Refer to https://kubernetes.io/docs/concepts/storage/volumes/
      volumeMounts:
        - mountPath: /path/to/log
          name: log-volume

    # the pod replicas in this group typed worker
    - replicas: 1
      # logical group name
      groupName: medium-group
      # pod type
      type: worker

      # custom labels. NOTE: do not define custom labels start with `raycluster.`, they may be used in controller.
      # Refer to https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
      labels:
        raycluster.group.name: medium-group

      # annotations for pod
      annotations:
        key: value

      # Command to start ray
      command: ray start --block --node-ip-address=$MY_POD_IP --address=$CLUSTER_NAME-head:6379 --object-manager-port=12345 --node-manager-port=12346 --object-store-memory=100000000 --num-cpus=1

      # use affinity to select nodes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: key1
                    operator: In
                    values: ["true"]
                  - key: key2
                    operator: In
                    values: ["true"]

      # use tolerations to select nodes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
      tolerations:
        - key: "key1"
          operator: "Equal"
          value: "true"
          effect: "NoSchedule"

      # resource requirements.
      # Refer to https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
      resources:
        limits:
          cpu: 2000m
          memory: 4Gi
          ephemeral-storage: 4Gi
        requests:
          cpu: 2000m
          memory: 4Gi
          ephemeral-storage: 4Gi

      # environment variables to set in the container.Optional.
      # Refer to https://kubernetes.io/docs/tasks/inject-data-application/define-environment-variable-container/
      containerEnv:
        - name: APPNAME
          value: raycluster-sample
        - name: MY_POD_NAME
          valueFrom:
            fieldRef:
              fieldPath: metadata.name
        - name: MY_POD_IP
          valueFrom:
            fieldRef:
              fieldPath: status.podIP

      # head service suffix: {namespace}.svc , follows Kubernetes standard
      headServiceSuffix: "ray-operator.svc"

      # use volumes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/storage/volumes/
      volumes:
        - name: log-volume
          emptyDir: {}

      # use volumeMounts.Optional.
      # Refer to https://kubernetes.io/docs/concepts/storage/volumes/
      volumeMounts:
        - mountPath: /path/to/log
          name: log-volume

    # the pod replicas in this group typed head
    - replicas: 1
      # logical group name
      groupName: headgroup
      # pod type
      type: head

      # custom labels. NOTE: do not define custom labels start with `raycluster.`, they may be used in controller.
      # Refer to https://kubernetes.io/docs/concepts/overview/working-with-objects/labels/
      labels:
        raycluster.group.name: headgroup

      # annotations for pod
      annotations:
        key: value

      # Command to start ray
      command: ray start --head --block --redis-port=6379 --node-ip-address=$MY_POD_IP --object-manager-port=12345 --node-manager-port=12346 --object-store-memory=100000000 --num-cpus=1

      # use affinity to select nodes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/configuration/assign-pod-node/#affinity-and-anti-affinity
      affinity:
        nodeAffinity:
          requiredDuringSchedulingIgnoredDuringExecution:
            nodeSelectorTerms:
              - matchExpressions:
                  - key: key1
                    operator: In
                    values: ["true"]
                  - key: key2
                    operator: In
                    values: ["true"]

      # use tolerations to select nodes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/configuration/taint-and-toleration/
      tolerations:
        - key: "key1"
          operator: "Equal"
          value: "true"
          effect: "NoSchedule"

      # resource requirements.
      # Refer to https://kubernetes.io/docs/concepts/configuration/manage-compute-resources-container/
      resources:
        limits:
          cpu: 1000m
          memory: 1Gi
          ephemeral-storage: 4Gi
        requests:
          cpu: 1000m
          memory: 1Gi
          ephemeral-storage: 4Gi

      # environment variables to set in the container.Optional.
      # Refer to https://kubernetes.io/docs/tasks/inject-data-application/define-environment-variable-container/
      containerEnv:
        - name: APPNAME
          value: raycluster-sample
        - name: MY_POD_NAME
          valueFrom:
            fieldRef:
              fieldPath: metadata.name
        - name: MY_POD_IP
          valueFrom:
            fieldRef:
              fieldPath: status.podIP

      # head service suffix: {namespace}.svc , follows Kubernetes standard
      headServiceSuffix: "ray-operator.svc"

      # use volumes.Optional.
      # Refer to https://kubernetes.io/docs/concepts/storage/volumes/
      volumes:
        - name: log-volume
          emptyDir: {}

      # use volumeMounts.Optional.
      # Refer to https://kubernetes.io/docs/concepts/storage/volumes/
      volumeMounts:
        - mountPath: /path/to/log
          name: log-volume