Kubernetes Restart a Pod by Deleting It

arazzo: 1.0.1
info:
  title: Kubernetes Restart a Pod by Deleting It
  summary: Confirm a pod is controller-owned, delete it gracefully, and watch the replacement appear.
  description: >-
    Kubernetes has no restart verb — the idiomatic restart is to delete the pod
    and let its owning controller recreate it. The workflow reads the pod first
    to confirm it carries an ownerReference, because deleting an unowned pod
    destroys it permanently rather than restarting it, then deletes it with an
    explicit grace period and lists the pods to observe the replacement. 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: workloadsApi
  url: ../openapi/kubernetes-workloads-api-openapi.yml
  type: openapi
workflows:
- workflowId: restart-pod
  summary: Delete a controller-owned pod so its controller recreates it.
  description: >-
    Reads the target pod and captures its ownerReferences, deletes it with a
    graceful termination period, and lists the namespace's matching pods so the
    freshly scheduled replacement can be observed.
  inputs:
    type: object
    required:
    - namespace
    - podName
    properties:
      namespace:
        type: string
        description: The namespace containing the pod.
      podName:
        type: string
        description: The name of the pod to restart.
      gracePeriodSeconds:
        type: integer
        description: Seconds to wait before forcefully terminating the pod.
      labelSelector:
        type: string
        description: Label selector matching the pod's siblings (e.g. app=nginx).
  steps:
  - stepId: readPod
    description: >-
      Read the pod and capture its ownerReferences. A pod with no owner will
      not be recreated after deletion, so this read is the safety check.
    operationId: getNamespacedPod
    parameters:
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: name
      in: path
      value: $inputs.podName
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      podName: $response.body#/metadata/name
      podUid: $response.body#/metadata/uid
      ownerReferences: $response.body#/metadata/ownerReferences
      previousPhase: $response.body#/status/phase
      nodeName: $response.body#/spec/nodeName
  - stepId: deletePod
    description: >-
      Delete the pod. The kubelet sends SIGTERM and waits the grace period
      before forcing termination; the owning controller schedules a replacement.
    operationId: deleteNamespacedPod
    parameters:
    - name: namespace
      in: path
      value: $inputs.namespace
    - name: name
      in: path
      value: $steps.readPod.outputs.podName
    - name: gracePeriodSeconds
      in: query
      value: $inputs.gracePeriodSeconds
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      deletionTimestamp: $response.body#/metadata/deletionTimestamp
      deletedPodName: $response.body#/metadata/name
  - stepId: observeReplacement
    description: >-
      List the pods matching the same label selector to observe the replacement
      pod. A new uid distinguishes the replacement from the deleted pod.
    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:
      pods: $response.body#/items
      replacementPodName: $response.body#/items/0/metadata/name
      replacementPhase: $response.body#/items/0/status/phase
  outputs:
    deletedPodName: $steps.deletePod.outputs.deletedPodName
    deletionTimestamp: $steps.deletePod.outputs.deletionTimestamp
    replacementPodName: $steps.observeReplacement.outputs.replacementPodName
    pods: $steps.observeReplacement.outputs.pods