arazzo: 1.0.1
info:
  title: Elasticsearch Cluster Health Triage
  summary: Walk cluster health, the cat health and nodes views, cluster stats, and cluster state to triage a degraded cluster.
  description: >-
    The operational sweep an on-call engineer runs when a cluster goes yellow or
    red. The workflow reads the structured health summary first, then widens out
    through the human-readable cat views for health and node membership, then
    pulls the aggregate stats that reveal resource pressure, and finally reads
    the full cluster state — the heaviest call, deliberately placed last so the
    cheap signals get a chance to explain the problem before it is made. 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: elasticsearchApi
  url: ../openapi/elasticsearch-openapi.yml
  type: openapi
workflows:
- workflowId: cluster-health-triage
  summary: Collect a full diagnostic picture of an Elasticsearch cluster, cheapest signal first.
  description: >-
    Reads cluster health, cat health, cat nodes, cluster stats, and cluster state
    in escalating order of cost, returning the status, unassigned shard count,
    node roster, and resource totals needed to triage a degraded cluster.
  inputs:
    type: object
    properties:
      verbose:
        type: boolean
        description: Whether the cat indices style views should include column headers.
  steps:
  - stepId: readHealth
    description: >-
      Read the structured cluster health summary. The status colour and the
      unassigned shard count are usually enough to classify the incident.
    operationId: clusterHealth
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      clusterName: $response.body#/cluster_name
      status: $response.body#/status
      numberOfNodes: $response.body#/number_of_nodes
      activeShards: $response.body#/active_shards
      unassignedShards: $response.body#/unassigned_shards
  - stepId: readCatHealth
    description: >-
      Read the cat health view for the same signal in the tabular form that gets
      pasted into an incident channel.
    operationId: catHealth
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      healthTable: $response.body
  - stepId: readCatNodes
    description: >-
      Read the node roster to confirm every expected node is still a member of
      the cluster, which is the most common explanation for unassigned shards.
    operationId: catNodes
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      nodesTable: $response.body
  - stepId: readClusterStats
    description: >-
      Pull aggregate cluster statistics for the index count, document totals, and
      store and heap pressure behind a degraded status.
    operationId: clusterStats
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      indexCount: $response.body#/indices/count
      documentCount: $response.body#/indices/docs/count
      storeSizeInBytes: $response.body#/indices/store/size_in_bytes
      nodeCount: $response.body#/nodes/count/total
  - stepId: readClusterState
    description: >-
      Read the full cluster state last. It is the most expensive call in the
      sweep, so it runs only after the cheaper signals have been captured.
    operationId: clusterState
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      stateVersion: $response.body#/version
      masterNode: $response.body#/master_node
  outputs:
    status: $steps.readHealth.outputs.status
    unassignedShards: $steps.readHealth.outputs.unassignedShards
    nodesTable: $steps.readCatNodes.outputs.nodesTable
    documentCount: $steps.readClusterStats.outputs.documentCount
    masterNode: $steps.readClusterState.outputs.masterNode