arazzo: 1.0.1
info:
  title: Kubernetes Retire an Application from a Namespace
  summary: Find an application's deployment, remove its service and deployment, and confirm the pods are gone.
  description: >-
    Removing one application from a shared namespace, where deleting the whole
    namespace is not an option because other workloads live there. The workflow
    locates the deployment, deletes the fronting service first so traffic stops
    arriving before the backends disappear, deletes the deployment so its
    ReplicaSet and pods are garbage-collected, and lists the pods to confirm
    termination. Every step spells out its request inline so the flow can be
    read and executed without opening the underlying OpenAPI description.
  version: 1.0.0
sourceDescriptions:
- name: kubernetesApi
  url: ../openapi/kubernetes-api-openapi.yml
  type: openapi
workflows:
- workflowId: retire-application
  summary: Delete an application's service and deployment from a shared namespace.
  description: >-
    Lists the namespace's deployments to locate the target, reads it for a
    record of what is being retired, deletes the service and then the
    deployment, and lists the remaining pods to confirm the cascade.
  inputs:
    type: object
    required:
    - namespace
    - deploymentName
    - serviceName
    properties:
      namespace:
        type: string
        description: The shared namespace the application lives in.
      deploymentName:
        type: string
        description: The deployment to retire.
      serviceName:
        type: string
        description: The service fronting the deployment, deleted before the backends.
      labelSelector:
        type: string
        description: Label selector matching the application's resources (e.g. app=nginx).
  steps:
  - stepId: findDeployment
    description: >-
      List the namespace's deployments narrowed to the application's label so
      the retirement targets only the intended workload.
    operationId: listNamespacedDeployments
    parameters:
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: labelSelector
      in: query
      value: $inputs.labelSelector
    - name: limit
      in: query
      value: 100
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      deployments: $response.body#/items
    onSuccess:
    - name: deploymentFound
      type: goto
      stepId: readDeployment
      criteria:
      - context: $response.body
        condition: $.items.length > 0
        type: jsonpath
    - name: nothingToRetire
      type: end
      criteria:
      - context: $response.body
        condition: $.items.length == 0
        type: jsonpath
  - stepId: readDeployment
    description: >-
      Read the deployment for a record of the image and replica count being
      retired, which is what a rollback would need to recreate it.
    operationId: getNamespacedDeployment
    parameters:
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: name
      in: path
      value: $inputs.deploymentName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      deploymentName: $response.body#/metadata/name
      retiredReplicas: $response.body#/spec/replicas
      retiredImage: $response.body#/spec/template/spec/containers/0/image
  - stepId: deleteService
    description: >-
      Delete the fronting service first so traffic stops being routed before
      the backends go away. A LoadBalancer service also deprovisions its
      cloud load balancer here.
    operationId: deleteNamespacedService
    parameters:
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: name
      in: path
      value: $inputs.serviceName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      deletedServiceName: $response.body#/metadata/name
  - stepId: deleteDeployment
    description: >-
      Delete the deployment. Its ReplicaSet and pods are garbage-collected via
      their ownerReferences and terminate gracefully.
    operationId: deleteNamespacedDeployment
    parameters:
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: name
      in: path
      value: $steps.readDeployment.outputs.deploymentName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      deletedDeploymentName: $response.body#/metadata/name
      deletionTimestamp: $response.body#/metadata/deletionTimestamp
  - stepId: confirmPodsGone
    description: >-
      List the pods still carrying the application's label to confirm the
      cascade removed them. Pods may briefly remain while terminating.
    operationId: listNamespacedPods
    parameters:
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: labelSelector
      in: query
      value: $inputs.labelSelector
    - name: limit
      in: query
      value: 100
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      remainingPods: $response.body#/items
  outputs:
    retiredImage: $steps.readDeployment.outputs.retiredImage
    retiredReplicas: $steps.readDeployment.outputs.retiredReplicas
    deletedServiceName: $steps.deleteService.outputs.deletedServiceName
    deletedDeploymentName: $steps.deleteDeployment.outputs.deletedDeploymentName
    remainingPods: $steps.confirmPodsGone.outputs.remainingPods