Prometheus Discover and Query a Metric

arazzo: 1.0.1
info:
  title: Prometheus Discover and Query a Metric
  summary: Walk from "what metrics exist?" to an evaluated PromQL result.
  description: >-
    The onboarding path every new Prometheus integrator takes. The workflow
    lists the metric names present in the TSDB, reads the type and help text for
    the metric of interest, enumerates the label names that metric carries,
    resolves the concrete series behind a selector, and finally evaluates a
    PromQL expression against it. 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: metadataApi
  url: ../openapi/prometheus-metadata-api-openapi.yml
  type: openapi
- name: queryApi
  url: ../openapi/prometheus-query-api-openapi.yml
  type: openapi
workflows:
- workflowId: discover-and-query-metric
  summary: Discover a metric, inspect its metadata and labels, then evaluate it.
  description: >-
    Uses the metadata endpoints to learn the shape of a metric before spending a
    query on it, which is the cheapest way to avoid writing a PromQL expression
    against a metric name or label that does not exist.
  inputs:
    type: object
    required:
    - metricName
    - seriesSelector
    - query
    properties:
      metricName:
        type: string
        description: The metric name to inspect (e.g. http_requests_total).
      seriesSelector:
        type: string
        description: >-
          A series selector used to resolve concrete series (e.g.
          http_requests_total{job="api"}).
      query:
        type: string
        description: >-
          The PromQL expression to evaluate (e.g. rate(http_requests_total[5m])).
      evaluationTime:
        type: string
        description: >-
          Optional evaluation timestamp as a Unix timestamp or RFC3339 string.
          Defaults to current server time when omitted.
  steps:
  - stepId: listMetricNames
    description: >-
      List every metric name in the TSDB by reading the values of the special
      __name__ label. This is how query UIs build metric autocomplete.
    operationId: getLabelValues
    parameters:
    - name: label_name
      in: path
      value: __name__
    - name: limit
      in: query
      value: 5000
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/status == "success"
    outputs:
      metricNames: $response.body#/data
  - stepId: describeMetric
    description: >-
      Read the metric type (counter, gauge, histogram, summary) and help string
      scraped from the targets exposing this metric. The type decides whether
      rate(), increase(), or a bare selector is the correct query.
    operationId: getMetricMetadata
    parameters:
    - name: metric
      in: query
      value: $inputs.metricName
    - name: limit_per_metric
      in: query
      value: 1
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/status == "success"
    outputs:
      metadata: $response.body#/data
  - stepId: listLabelNames
    description: >-
      List the label names actually present on this metric, restricted by the
      series selector so the result is scoped to the metric under investigation.
    operationId: getLabelNames
    parameters:
    - name: "match[]"
      in: query
      value:
      - $inputs.seriesSelector
    - name: limit
      in: query
      value: 500
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/status == "success"
    outputs:
      labelNames: $response.body#/data
  - stepId: resolveSeries
    description: >-
      Resolve the selector to the concrete label sets it matches. An empty result
      here means the selector matches nothing and any query built on it will
      return no data.
    operationId: querySeries
    parameters:
    - name: "match[]"
      in: query
      value:
      - $inputs.seriesSelector
    - name: limit
      in: query
      value: 100
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/status == "success"
    outputs:
      seriesLabelSets: $response.body#/data
      firstSeries: $response.body#/data/0
  - stepId: evaluateExpression
    description: >-
      Evaluate the PromQL expression at a single point in time and return the
      resulting vector.
    operationId: queryInstant
    parameters:
    - name: query
      in: query
      value: $inputs.query
    - name: time
      in: query
      value: $inputs.evaluationTime
    - name: timeout
      in: query
      value: 30s
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/status == "success"
    outputs:
      resultType: $response.body#/data/resultType
      result: $response.body#/data/result
      firstValue: $response.body#/data/result/0/value/1
  outputs:
    metricNames: $steps.listMetricNames.outputs.metricNames
    metadata: $steps.describeMetric.outputs.metadata
    labelNames: $steps.listLabelNames.outputs.labelNames
    seriesLabelSets: $steps.resolveSeries.outputs.seriesLabelSets
    resultType: $steps.evaluateExpression.outputs.resultType
    result: $steps.evaluateExpression.outputs.result