Prometheus Manage a Grouped Metric Group End to End

arazzo: 1.0.1
info:
  title: Prometheus Manage a Grouped Metric Group End to End
  summary: Push, update, verify, and tear down metrics scoped by instance-level grouping labels.
  description: >-
    When several workers run the same job concurrently, pushing on the job label
    alone makes them overwrite each other. Grouping labels give each worker its
    own group, addressed as alternating label-name and label-value path
    segments. The workflow replaces one worker's group, updates a single metric
    within it, verifies the stored state, and deletes the group when the worker
    is torn down. 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: pushgatewayApi
  url: ../openapi/prometheus-pushgateway-api-openapi.yml
  type: openapi
workflows:
- workflowId: push-grouped-metrics-lifecycle
  summary: Full lifecycle of a metric group identified by job plus grouping labels.
  description: >-
    The grouping labels form part of the group's identity, so the delete at the
    end must carry exactly the same label path as the pushes, or it silently
    removes nothing.
  inputs:
    type: object
    required:
    - jobName
    - groupingLabels
    - metricsPayload
    - updatePayload
    properties:
      jobName:
        type: string
        description: The job label for this metric group (e.g. batch-importer).
      groupingLabels:
        type: string
        description: >-
          Additional grouping labels as alternating label-name/label-value path
          segments (e.g. instance/server01 or instance/host1/datacenter/eu).
      metricsPayload:
        type: string
        description: >-
          The worker's metrics in Prometheus text exposition format, with a
          trailing newline.
      updatePayload:
        type: string
        description: >-
          A subset of metrics in text exposition format to update within the
          existing group without replacing it.
  steps:
  - stepId: replaceGroupedMetrics
    description: >-
      Atomically replace this worker's group. Because the group is keyed on job
      plus the grouping labels, this touches only this worker and leaves every
      sibling worker's group intact.
    operationId: replaceGroupedMetrics
    parameters:
    - name: job
      in: path
      value: $inputs.jobName
    - name: labels
      in: path
      value: $inputs.groupingLabels
    requestBody:
      contentType: text/plain
      payload: $inputs.metricsPayload
    successCriteria:
    - condition: $statusCode == 200
  - stepId: updateGroupedMetrics
    description: >-
      Update a subset of the metrics in the group. POST preserves the metrics
      already in the group, which is what makes progress reporting from a
      long-running worker possible without re-sending everything.
    operationId: pushGroupedMetrics
    parameters:
    - name: job
      in: path
      value: $inputs.jobName
    - name: labels
      in: path
      value: $inputs.groupingLabels
    requestBody:
      contentType: text/plain
      payload: $inputs.updatePayload
    successCriteria:
    - condition: $statusCode == 200
  - stepId: verifyGroupState
    description: >-
      Read back the stored groups and confirm this worker's group carries both
      the replaced metrics and the update, under the expected label set.
    operationId: listAllMetrics
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/status == "success"
    outputs:
      metricGroups: $response.body#/data
      firstGroupLabels: $response.body#/data/0/labels
      firstGroupLastPushedAt: $response.body#/data/0/lastPushedAt
      firstGroupMetricFamilies: $response.body#/data/0/metricFamilies
  - stepId: deleteGroupedMetrics
    description: >-
      Delete this worker's group at teardown. Only the group with this exact
      label set is removed. Skipping this is what leaves a decommissioned
      worker's last values being scraped indefinitely as though they were live.
    operationId: deleteGroupedMetrics
    parameters:
    - name: job
      in: path
      value: $inputs.jobName
    - name: labels
      in: path
      value: $inputs.groupingLabels
    successCriteria:
    - condition: $statusCode == 202
  outputs:
    metricGroups: $steps.verifyGroupState.outputs.metricGroups
    groupLabels: $steps.verifyGroupState.outputs.firstGroupLabels
    lastPushedAt: $steps.verifyGroupState.outputs.firstGroupLastPushedAt