deployment-config-custom-strategy

Objectives

• To show how to create a Custom Strategy with your deployment logic using Openshift' DeploymentConfig.
• To Show deployment Strategies modes using deploymentConfig that has no downtime at all, and favouring new version replicas all the time over the older version while the deployment is not finished.

Demos

Create new project:

oc new-project dc-test-custom-strategy

Custom Strategy Using DeploymentConfig.

1. Need to create RBAC Permissions for the deployer ServiceAccount in the project, so it will be able to get/list DeploymentConfig Resource

Role.yaml

apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  name: get-deployment-config
rules:
  - apiGroups:
      - apps.openshift.io
    resources:
      - deploymentconfigs
    verbs:
      - get
      - list
      - watch
  - apiGroups:
      - "apps.openshift.io"
    resources:
      - deploymentconfigs/scale
    verbs:
      - get
      - update

RoleBinding.yaml:

apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: enrich-deployer-with-get-dc
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: get-deployment-config
subjects:
  - kind: ServiceAccount
    name: deployer
    namespace: dc-test-custom-strategy

2. Create DeploymentConfig resource with Custom Strategy:

apiVersion: apps.openshift.io/v1
kind: DeploymentConfig
metadata:
  name: dc-custom-strategy-test
spec:
  replicas: 4
  selector:
    deployment-config.name: dc-custom-strategy-test
  strategy:
    type: Recreate
    customParams:
      image: openshift/origin-deployer
      command:
      - /usr/bin/bash
      - -c
      - |
        set -e
        export DC_NAME=$(oc get rc ${OPENSHIFT_DEPLOYMENT_NAME} -o=jsonpath="{.metadata.ownerReferences[0].name}" -n ${OPENSHIFT_DEPLOYMENT_NAMESPACE})
        export REPLICAS=$(oc get deploymentconfig ${DC_NAME} -o jsonpath="{.spec.replicas}" -n ${OPENSHIFT_DEPLOYMENT_NAMESPACE})
        echo "About to scale down to 1"
        oc scale rc/${OPENSHIFT_DEPLOYMENT_NAME} --replicas=1 -n ${OPENSHIFT_DEPLOYMENT_NAMESPACE}
        readyReplicas=$(oc get rc ${OPENSHIFT_DEPLOYMENT_NAME} -o=jsonpath="{.status.readyReplicas}")
        echo "Waiting for 1 replica of new version to be ready"
        while [ "$readyReplicas" == "" ] 
        do
          sleep 1
          readyReplicas=$(oc get rc ${OPENSHIFT_DEPLOYMENT_NAME} -n ${OPENSHIFT_DEPLOYMENT_NAMESPACE} -o=jsonpath="{.status.readyReplicas}")
        done
        echo "New version Replica is Ready"
        echo "Continuing with the replacing all older version replicas with new ones at once!" 
        openshift-deploy --deployment ${OPENSHIFT_DEPLOYMENT_NAME} --until 100%

  template:
    metadata:
      labels:
        deployment-config.name: dc-custom-strategy-test
        app-version: v1
    spec:
      containers:
      - image: busybox
        name: default-container
        imagePullPolicy: IfNotPresent
        command:
        - /bin/sh
        - -c
        - "sleep 15 ; echo \"service is ready\" ; echo 1  >> /tmp/readiness.out ; sleep infinity"
        readinessProbe:
          initialDelaySeconds: 5
          periodSeconds: 5
          failureThreshold: 4
          exec:
            command:
            - cat
            - /tmp/readiness.out

  test: false
  triggers:
    - type: ConfigChange

3. Now apply using a kustomization overlay all the above to the new project:

oc kustomize .  | oc apply -f -

4. Watch the deployment process in live

watch --interval 1 oc get pods  --sort-by '{.metadata.name}'

5. When all replicas are ready, change .spec.template.metadata.labels.app-version (for example , to value of v2) in deploymentconfig manifest file , and apply again deploymentconfig to cluster:

oc apply -f deployment-config.yaml

Or directly with json patch

oc patch dc dc-custom-strategy-test  --type='json' -p='[{"op": "replace", "path": "/spec/template/metadata/labels/app-version", "value": "v2" }]'

6. Watch the deployment process in live, take a look how no downtime of service is achieved.

watch --interval 1 oc get pods  --sort-by '{.metadata.name}

7. When deployment is over, delete everything:

kustomize build . | oc delete -f -

Demo:

Rolling Update Strategy With Specific RollingUpdate Params

There is a formula that should do the work in general , if you consider the number of replicas to be n , then you define it as follows:

for n=1 , it's trivial and you don't need to set any special parameters, just use rolling update strategy, which is the default.

for n > 1 , you should define the following:

• spec.strategy.rollingParams.maxSurge = 1

• spec.strategy.rollingParams.maxUnavailable = n-1

1. For example, for n=4 , Need to deploy the following deployment config:

apiVersion: apps.openshift.io/v1
kind: DeploymentConfig
metadata:
  name: dc-rolling-strategy-test
spec:
  replicas: 4
  selector:
    deployment-config.name: dc-custom-strategy-test
  strategy:
    type: Rolling
    customParams:
      image: openshift/origin-deployer
    rollingParams:
      maxSurge: 1
      maxUnavailable: 3
  template:
    metadata:
      labels:
        deployment-config.name: dc-custom-strategy-test
        app-version: v1
    spec:
      containers:
      - image: busybox
        name: default-container
        imagePullPolicy: IfNotPresent
        command:
        - /bin/sh
        - -c
        - "sleep 15 ; echo \"service is ready\" ; echo 1  >> /tmp/readiness.out ; sleep infinity"
        readinessProbe:
          initialDelaySeconds: 5
          periodSeconds: 5
          failureThreshold: 4
          exec:
            command:
            - cat
            - /tmp/readiness.out

  test: false
  triggers:
    - type: ConfigChange

oc apply -f using-rolling-update-parameters/deployment-config.yaml

2. Watch the deployment process in live

watch --interval 1 oc get pods  --sort-by '{.metadata.name}'

3. Once all replicas are up and ready, "upgrade" to next version by using a json patch:

oc patch dc dc-rolling-strategy-test  --type='json' -p='[{"op": "replace", "path": "/spec/template/metadata/labels/app-version", "value": "v2" }]'

4. Watch the deployment process in live, take a look how no downtime of service is achieved.

watch --interval 1 oc get pods  --sort-by '{.metadata.name}

5. When all pods of new revision is up and ready, you can delete it to release resources:

oc delete -f using-rolling-update-parameters/deployment-config.yaml