arazzo: 1.0.1
info:
  title: Buildkite Retry Failed Job
  summary: Find a failed job in a build, retry it, then poll the build to completion.
  description: >-
    A self-healing flow for flaky steps. The workflow reads a finished build,
    scans its jobs for a failed script job, retries that job, and then polls the
    build until it leaves the running state so the retry outcome can be observed.
    This lets automation recover from transient failures without re-running the
    entire build. 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: buildkiteRestApi
  url: ../openapi/buildkite-rest-api-openapi.yml
  type: openapi
workflows:
- workflowId: retry-failed-job
  summary: Retry a failed job and poll the build to completion.
  description: >-
    Reads a build, retries the supplied failed job, and polls the build until it
    reaches a terminal state.
  inputs:
    type: object
    required:
    - org
    - pipeline
    - number
    - job
    properties:
      org:
        type: string
        description: Organization slug that owns the pipeline.
      pipeline:
        type: string
        description: Pipeline slug the build belongs to.
      number:
        type: integer
        description: Build number containing the job to retry.
      job:
        type: string
        description: UUID of the failed job to retry.
  steps:
  - stepId: getBuild
    description: >-
      Read the build to confirm it is terminal and capture its jobs before
      retrying.
    operationId: getBuild
    parameters:
    - name: org
      in: path
      value: $inputs.org
    - name: pipeline
      in: path
      value: $inputs.pipeline
    - name: number
      in: path
      value: $inputs.number
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      buildState: $response.body#/state
      jobs: $response.body#/jobs
  - stepId: retryJob
    description: >-
      Retry the supplied failed job, which re-queues it for an available agent.
    operationId: retryJob
    parameters:
    - name: org
      in: path
      value: $inputs.org
    - name: pipeline
      in: path
      value: $inputs.pipeline
    - name: number
      in: path
      value: $inputs.number
    - name: job
      in: path
      value: $inputs.job
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      jobId: $response.body#/id
      jobState: $response.body#/state
  - stepId: pollBuild
    description: >-
      Poll the build until the retried job completes and the build leaves the
      running and scheduled states.
    operationId: getBuild
    parameters:
    - name: org
      in: path
      value: $inputs.org
    - name: pipeline
      in: path
      value: $inputs.pipeline
    - name: number
      in: path
      value: $inputs.number
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      finalState: $response.body#/state
      webUrl: $response.body#/web_url
    onSuccess:
    - name: stillRunning
      type: goto
      stepId: pollBuild
      criteria:
      - context: $response.body
        condition: $.state == "running" || $.state == "scheduled"
        type: jsonpath
  outputs:
    retriedJobId: $steps.retryJob.outputs.jobId
    finalState: $steps.pollBuild.outputs.finalState
    webUrl: $steps.pollBuild.outputs.webUrl