Prometheus Silence a Firing Alert

arazzo: 1.0.1
info:
  title: Prometheus Silence a Firing Alert
  summary: Find a firing alert, silence it by its labels, and confirm the silence took hold.
  description: >-
    The standard incident-response action: an alert is firing, the cause is
    known and being worked, and the page needs to stop without touching the
    rule. The workflow locates the firing alert to harvest its real labels,
    creates a time-boxed silence whose matchers target those labels, reads the
    silence back to confirm it is active rather than pending, and then re-reads
    the alert to prove Alertmanager now reports it suppressed. 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: alertmanagerApi
  url: ../openapi/prometheus-alertmanager-api-openapi.yml
  type: openapi
workflows:
- workflowId: silence-firing-alert
  summary: Create a silence for a firing alert and verify it is suppressed.
  description: >-
    Reading the alert first matters — silences match on labels, and a matcher
    written from memory rather than from the alert's actual label set is the
    usual reason a silence appears to do nothing.
  inputs:
    type: object
    required:
    - alertName
    - alertNameMatcher
    - instance
    - startsAt
    - endsAt
    - createdBy
    - comment
    properties:
      alertName:
        type: string
        description: The alertname label value to silence (e.g. HighRequestLatency).
      alertNameMatcher:
        type: string
        description: >-
          The alert expressed as a label matcher for filtering (e.g.
          alertname="HighRequestLatency").
      instance:
        type: string
        description: >-
          The instance label value to scope the silence to a single target
          rather than every instance of the alert.
      startsAt:
        type: string
        description: When the silence takes effect, as an RFC3339 timestamp.
      endsAt:
        type: string
        description: >-
          When the silence expires, as an RFC3339 timestamp. Always set this to
          a real bound so the silence cannot outlive the incident.
      createdBy:
        type: string
        description: The person or system creating the silence.
      comment:
        type: string
        description: The reason for the silence, including a ticket reference.
  steps:
  - stepId: findFiringAlert
    description: >-
      Locate the firing alert and capture the label set Alertmanager is actually
      holding for it. The fingerprint identifies this unique label combination.
    operationId: getAlerts
    parameters:
    - name: filter
      in: query
      value:
      - $inputs.alertNameMatcher
    - name: active
      in: query
      value: true
    - name: silenced
      in: query
      value: false
    - name: inhibited
      in: query
      value: false
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      alerts: $response.body
      fingerprint: $response.body#/0/fingerprint
      labels: $response.body#/0/labels
      startedAt: $response.body#/0/startsAt
  - stepId: listReceivers
    description: >-
      List the configured receivers so the operator can record which
      notification integrations this silence is about to mute.
    operationId: listReceivers
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      receivers: $response.body
      firstReceiverName: $response.body#/0/name
  - stepId: createSilence
    description: >-
      Create the silence. Every matcher must match for an alert to be
      suppressed, so alertname plus instance scopes this to the one firing
      target and leaves the same alert on other instances still paging.
    operationId: createSilence
    requestBody:
      contentType: application/json
      payload:
        matchers:
        - name: alertname
          value: $inputs.alertName
          isRegex: false
          isEqual: true
        - name: instance
          value: $inputs.instance
          isRegex: false
          isEqual: true
        startsAt: $inputs.startsAt
        endsAt: $inputs.endsAt
        createdBy: $inputs.createdBy
        comment: $inputs.comment
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      silenceID: $response.body#/silenceID
  - stepId: confirmSilence
    description: >-
      Read the silence back by its id and check its state. A state of pending
      means startsAt is in the future and the alert will keep paging until then.
    operationId: getSilence
    parameters:
    - name: silenceID
      in: path
      value: $steps.createSilence.outputs.silenceID
    successCriteria:
    - condition: $statusCode == 200
    - condition: $response.body#/status/state == "active"
    outputs:
      silenceId: $response.body#/id
      state: $response.body#/status/state
      matchers: $response.body#/matchers
      expiresAt: $response.body#/endsAt
  - stepId: verifyAlertSuppressed
    description: >-
      Re-read the alert including silenced alerts and confirm Alertmanager now
      reports it suppressed and attributes the suppression to this silence.
    operationId: getAlerts
    parameters:
    - name: filter
      in: query
      value:
      - $inputs.alertNameMatcher
    - name: active
      in: query
      value: false
    - name: silenced
      in: query
      value: true
    - name: inhibited
      in: query
      value: false
    successCriteria:
    - condition: $statusCode == 200
    outputs:
      suppressedAlerts: $response.body
      statusState: $response.body#/0/status/state
      silencedBy: $response.body#/0/status/silencedBy
  outputs:
    fingerprint: $steps.findFiringAlert.outputs.fingerprint
    silenceID: $steps.createSilence.outputs.silenceID
    state: $steps.confirmSilence.outputs.state
    expiresAt: $steps.confirmSilence.outputs.expiresAt
    silencedBy: $steps.verifyAlertSuppressed.outputs.silencedBy