Elastic Container Instance:Schedule pods to ARM-based virtual nodes

Configure the nodeSelector field

ARM-based virtual nodes in ACK clusters and ACK Serverless clusters have the kubernetes.io/arch: arm64 label. You can use this label in a nodeSelector to schedule an ECI Pod to an ARM-based virtual node.

nodeSelector:
  kubernetes.io/arch: arm64 

Sample YAML file of a pod:

• No specifications are specified for the pod.

  If you do not specify an ECS instance type, the system defaults to ecs.c8y.large, an ARM-based instance type with 2 vCPUs and 4 GiB of memory.

  apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: test
    labels:
      app: test
  spec:
    replicas: 1
    selector:
      matchLabels:
        app: nginx
    template:
      metadata:
        name: nginx-test
        labels:
          app: nginx
          alibabacloud.com/eci: "true"    # Schedule the Pod to a virtual node.
      spec:
        containers:
        - name: nginx
          image: arm64v8/centos:7.9.2009   # Use an image based on the ARM architecture.
          command: ["sleep"]
          args: ["999999"]
        nodeSelector:
          kubernetes.io/arch: arm64  # Specify an ARM-based node.

• vCPU and memory specifications are specified.

  If you specify vCPU and memory resources, the system automatically selects a matching ARM-based ECS instance type from the instance types supported by ECI.

  Note

  You can specify vCPU and memory by defining the limits or requests fields for the container (using limits is recommended), or by adding the k8s.aliyun.com/eci-use-specs annotation.

  apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: test
    labels:
      app: test
  spec:
    replicas: 1
    selector:
      matchLabels:
        app: nginx
    template:
      metadata:
        name: nginx-test
        labels:
          app: nginx
          alibabacloud.com/eci: "true"    # Schedule the Pod to a virtual node.
      spec:
        containers:
        - name: nginx
          image: arm64v8/centos:7.9.2009   # Use an image based on the ARM architecture.
          command: ["sleep"]
          args: ["999999"]
          resources:
            limits:
              cpu: "1000m"      # Specify 1 vCPU.
              memory: "4096Mi"   # Specify 4 GiB of memory for the nginx container.
        nodeSelector:
          kubernetes.io/arch: arm64  # Specify an ARM-based node.

• Specify an ARM-based ECS instance type

  If you have special requirements for the specifications, you can use the k8s.aliyun.com/eci-use-specs annotation to specify specific Arm specifications. For information about the Arm specifications that ECI supports, see Specify Arm specifications to create a pod.

  apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: test
    labels:
      app: test
  spec:
    replicas: 1
    selector:
      matchLabels:
        app: nginx
    template:
      metadata:
        name: nginx-test
        labels:
          app: nginx
          alibabacloud.com/eci: "true" 
        annotations:
          k8s.aliyun.com/eci-use-specs: "ecs.c8y.large,ecs.g8y.large"   # Specify the supported ARM-based ECS instance types. You can specify up to five types.
      spec:
        containers:
        - name: nginx
          image: arm64v8/centos:7.9.2009   # Use an image based on the ARM architecture.
          command: ["sleep"]
          args: ["999999"]
        nodeSelector:
          kubernetes.io/arch: arm64  # Schedule the Pod to an ARM-based node.

Configure tolerations and nodeaffinity to schedule multi-architecture images

If you use a multi-architecture image and need to schedule Pods across both x86 and ARM architectures, you can use tolerations and nodeAffinity to set a scheduling preference for either ARM-based or x86-based virtual nodes.

Sample YAML file of a pod:

• Prefer ARM-based nodes

  apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: test
    labels:
      app: test
  spec:
    replicas: 1
    selector:
      matchLabels:
        app: nginx
    template:
      metadata:
        name: nginx-test
        labels:
          app: nginx
          alibabacloud.com/eci: "true"    # Schedule the Pod to a virtual node.
      spec:
        tolerations:                 # Tolerate the taint on ARM-based nodes.
        - key: kubernetes.io/arch
          operator: Equal
          value: arm64
          effect: NoSchedule
        affinity:                 
          nodeAffinity:     
            preferredDuringSchedulingIgnoredDuringExecution:
            - weight: 1
              preference:
                matchExpressions:
                - key: kubernetes.io/arch
                  operator: In
                  values:
                  - arm64       # Prefer ARM-based nodes.
        containers:
        - name: nginx
          image: nginx                          # Use a multi-architecture image.
          command: ["sleep"]
          args: ["999999"]

• Preferentially schedule the pod to an x86-based virtual node

  apiVersion: apps/v1
  kind: Deployment
  metadata:
    name: test
    labels:
      app: test
  spec:
    replicas: 1
    selector:
      matchLabels:
        app: nginx
    template:
      metadata:
        name: nginx-test
        labels:
          app: nginx
          alibabacloud.com/eci: "true"    # Schedule the Pod to a virtual node.
      spec:
        tolerations:                 # Tolerate the taint on ARM-based nodes.
        - key: kubernetes.io/arch
          operator: Equal
          value: arm64
          effect: NoSchedule
        affinity:                 
          nodeAffinity:     
            preferredDuringSchedulingIgnoredDuringExecution:
            - weight: 1
              preference:
                matchExpressions:
                - key: kubernetes.io/arch
                  operator: In
                  values:
                  - amd64         # Prefer x86-based nodes.
        containers:
        - name: nginx
          image: nginx                          # Use a multi-architecture image.
          command: ["sleep"]
          args: ["999999"]