Docker Troubleshoot a Failing Container

arazzo: 1.0.1
info:
  title: Docker Troubleshoot a Failing Container
  summary: Gather the full diagnostic picture for a container — state, logs, processes, resource stats, and filesystem drift.
  description: >-
    The read-only flow an operator or an agent runs when a container is
    misbehaving. It collects the container's low-level state and exit code, the
    tail of its stdout and stderr, the process table inside it, a one-shot
    resource snapshot, and the list of files that have drifted from the image
    layer. Nothing here mutates the container, so the workflow is safe to run
    against production. 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: dockerEngineApi
  url: ../openapi/docker-openapi.yml
  type: openapi
workflows:
- workflowId: troubleshoot-container
  summary: Collect state, logs, processes, stats, and filesystem changes for one container.
  description: >-
    Reads the container's inspect payload for state and exit code, tails its
    logs, lists the processes running inside it, takes a single non-streaming
    resource sample, and reports which files have changed relative to the image.
  inputs:
    type: object
    required:
    - containerId
    properties:
      containerId:
        type: string
        description: ID or name of the container to diagnose.
      tail:
        type: string
        description: Number of log lines to return from the end of the logs, or "all".
      since:
        type: integer
        description: Only return logs since this UNIX timestamp.
  steps:
  - stepId: inspectState
    description: >-
      Read the container's low-level state, including whether it is running, its
      exit code, OOM-killed flag, restart count, and health status. This is the
      anchor for the whole diagnosis.
    operationId: ContainerInspect
    parameters:
    - name: id
      in: path
      value: $inputs.containerId
    - name: size
      in: query
      value: true
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      resolvedId: $response.body#/Id
      state: $response.body#/State
      restartCount: $response.body#/RestartCount
      hostConfig: $response.body#/HostConfig
      sizeRw: $response.body#/SizeRw
  - stepId: readLogs
    description: >-
      Tail both stdout and stderr with timestamps so the failure can be placed on
      a timeline against the state transitions above.
    operationId: ContainerLogs
    parameters:
    - name: id
      in: path
      value: $steps.inspectState.outputs.resolvedId
    - name: stdout
      in: query
      value: true
    - name: stderr
      in: query
      value: true
    - name: timestamps
      in: query
      value: true
    - name: follow
      in: query
      value: false
    - name: tail
      in: query
      value: $inputs.tail
    - name: since
      in: query
      value: $inputs.since
    successCriteria:
    - condition: $statusCode == 200
  - stepId: listProcesses
    description: >-
      List the processes running inside the container. An empty or unexpected
      process table often explains a container that is "running" but idle. This
      endpoint is not supported on Windows containers.
    operationId: ContainerTop
    parameters:
    - name: id
      in: path
      value: $steps.inspectState.outputs.resolvedId
    - name: ps_args
      in: query
      value: aux
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      titles: $response.body#/Titles
      processes: $response.body#/Processes
  - stepId: sampleStats
    description: >-
      Take a single resource sample rather than opening the default stats stream,
      so the workflow terminates. Memory and CPU here corroborate an OOM kill seen
      in the inspect state.
    operationId: ContainerStats
    parameters:
    - name: id
      in: path
      value: $steps.inspectState.outputs.resolvedId
    - name: stream
      in: query
      value: false
    - name: one-shot
      in: query
      value: true
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      memoryStats: $response.body#/memory_stats
      cpuStats: $response.body#/cpu_stats
      pidsStats: $response.body#/pids_stats
  - stepId: listFilesystemChanges
    description: >-
      Report which files have been added, modified, or deleted relative to the
      image layer. Unexpected writes here point at state that will be lost on
      recreate, or at a container being used as a pet.
    operationId: ContainerChanges
    parameters:
    - name: id
      in: path
      value: $steps.inspectState.outputs.resolvedId
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      changes: $response.body
  outputs:
    containerId: $steps.inspectState.outputs.resolvedId
    state: $steps.inspectState.outputs.state
    processes: $steps.listProcesses.outputs.processes
    memoryStats: $steps.sampleStats.outputs.memoryStats
    changes: $steps.listFilesystemChanges.outputs.changes