Binding a New EIP to a Pod

Assigning an EIP During Pod Creation

If you configure the pod-with-eip annotation when creating a pod, an EIP will be automatically assigned and bound to the pod.

The following uses a pod named nginx as an example. For details about the parameters, see Table 1.

• To create a pod with an EIP that uses a dedicated bandwidth, you do not need to specify the bandwidth ID.

  kind: Pod
  apiVersion: cci/v2
  metadata:
    name: nginx
    namespace: ns-test
    annotations:
      yangtse.io/pod-with-eip: "true"        # An EIP will be automatically assigned when the pod is being created.
      yangtse.io/eip-bandwidth-size: "5"     # EIP bandwidth.
      yangtse.io/eip-network-type: 5_bgp     # EIP type
      yangtse.io/eip-charge-mode: traffic  # EIP billing mode
      yangtse.io/eip-bandwidth-name: "xxx"   # Bandwidth name
  spec:
    containers:
      - name: container1
        image: nginx:stable-alpine-perl
        resources:
          limits:
            cpu: '1'
            memory: 2G
          requests:
            cpu: '1'
            memory: 2G
        terminationMessagePath: /dev/termination-log
        terminationMessagePolicy: File
    restartPolicy: Always
    terminationGracePeriodSeconds: 30
    dnsPolicy: Default
    securityContext: {}
    imagePullSecrets:
      - name: imagepull-secret

• To create a pod with an EIP that uses a shared bandwidth, you must and only need to specify the bandwidth ID.

   
  kind: Pod
  apiVersion: cci/v2
  metadata:
    name: nginx
    namespace: ns-test
    annotations:
      yangtse.io/pod-with-eip: "true"  # An EIP will be automatically assigned when the pod is being created.
      yangtse.io/eip-bandwidth-id: "5xxxxxx4-a801-4bfd-9004-5fxxxxxxxx19" # EIP shared bandwidth ID
  spec:
    containers:
      - name: container1
        image: nginx:stable-alpine-perl
        resources:
          limits:
            cpu: '1'
            memory: 2G
          requests:
            cpu: '1'
            memory: 2G
        terminationMessagePath: /dev/termination-log
        terminationMessagePolicy: File
    restartPolicy: Always
    terminationGracePeriodSeconds: 30
    dnsPolicy: Default
    securityContext: {}
    imagePullSecrets:
      - name: imagepull-secret

Verifying the EIP Readiness

The startup time of the pod may be earlier than the time when the EIP assignment result is returned. During pod startup, the EIP may fail to be bound.

You can use an init container to check whether the EIP is assigned. After the pod IP address is assigned, the container network controller binds an EIP to the pod and returns the assignment result to the yangtse.io/allocated-ipv4-eip annotation of the pod. You can configure an init container in the pod and use the downwardAPI to mount the annotation to the init container through a volume to verify that the EIP has been assigned. For details about the parameters, see Table 2. You can configure the init container as follows:

CCI allows EIPs to be automatically bound to pods. EIPs are assigned after pod scheduling is complete. EIP annotations cannot be injected to pods through ENV.

kind: Pod
apiVersion: cci/v2
metadata:
  name: nginx
  namespace: ns-test
  annotations:
    yangtse.io/pod-with-eip: "true"  
    yangtse.io/eip-bandwidth-size: "5" 
    yangtse.io/eip-network-type: 5_bgp
    yangtse.io/eip-charge-mode: traffic
    yangtse.io/eip-bandwidth-name: "xxx"
spec:
  initContainers:
  - name: init
    image: busybox:latest
    command: ['timeout', '60', 'sh', '-c', "until grep -E '[0-9]+' /etc/eipinfo/allocated-ipv4-eip; do echo waiting for allocated-ipv4-eip; sleep 2; done"]
    volumeMounts:
        - name: eipinfo
          mountPath: /etc/eipinfo
  volumes:
    - name: eipinfo
      downwardAPI:
        items:
          - path: "allocated-ipv4-eip"
            fieldRef:
              fieldPath: metadata.annotations['yangtse.io/allocated-ipv4-eip']
...

Releasing an EIP with a Pod

When you delete a pod, the EIP created with it is also released. You can run the following command to delete a pod using ccictl:

# ccictl delete pod $pod_name -n $namespace_name