Google Cloud Spanner High Priority CPU Utilization Sentinel

id: gcp-spanner-high-priority-cpu-saturation-sentinel
namespace: company.cloud

inputs:
  - id: project_id
    type: STRING
    description: Google Cloud Project ID.
  - id: instance_id
    type: STRING
    description: Google Cloud Spanner instance ID.
  - id: cpu_threshold_pct
    type: INT
    defaults: 65
    description: "Maximum allowable high-priority CPU utilization percentage before
      alerting (default 65%)."

tasks:
  - id: query_spanner_cpu
    type: io.kestra.plugin.gcp.cli.GCloudCLI
    projectId: "{{ inputs.project_id }}"
    serviceAccount: "{{ secret('GCP_SERVICE_ACCOUNT') }}"
    commands:
      - |
        FILTER='metric.type="spanner.googleapis.com/instance/cpu/utilization_by_priority" AND resource.labels.instance_id="{{ inputs.instance_id }}" AND metric.labels.priority="high"'
        START_TIME=$(date -u -v-15M +%Y-%m-%dT%H:%M:%SZ)

        gcloud monitoring timeseries list           --filter="$FILTER"           --interval-start-time="$START_TIME"           --format=json > /tmp/spanner_cpu.json

        PEAK_CPU=$(jq '[.[]?.points[]?.value.doubleValue // 0.0] | max // 0.0' /tmp/spanner_cpu.json | awk '{print int($1 * 100)}')
        echo "::set-output name=peak_cpu::$PEAK_CPU::"

  - id: evaluate_cpu_saturation
    type: io.kestra.plugin.core.flow.If
    condition: "{{ outputs.query_spanner_cpu.vars.peak_cpu | int >=
      inputs.cpu_threshold_pct }}"
    then:
      - id: notify_dba_team
        type: io.kestra.plugin.slack.notifications.SlackIncomingWebhook
        url: "{{ secret('SLACK_WEBHOOK') }}"
        payload: |
          {
            "channel": "#cloud-ops",
            "attachments": [
              {
                "color": "danger",
                "title": "🚨 Google Cloud Spanner High Priority CPU Saturated Alert",
                "text": "Cloud Spanner instance `{{ inputs.instance_id }}` in project `{{ inputs.project_id }}` reached peak high-priority CPU utilization of *{{ outputs.query_spanner_cpu.vars.peak_cpu }}%* (threshold {{ inputs.cpu_threshold_pct }}%)! Recommended maximum for regional instances (65%) or multi-region instances (45%) has been exceeded, threatening transactional latency SLA breaches.",
                "footer": "Kestra Cloud Sentinel • Google Cloud Spanner"
              }
            ]
          }

errors:
  - id: alert_pipeline_failure
    type: io.kestra.plugin.slack.notifications.SlackIncomingWebhook
    url: "{{ secret('SLACK_WEBHOOK') }}"
    payload: |
      {
        "channel": "#cloud-ops",
        "attachments": [
          {
            "color": "danger",
            "title": "⚠️ Google Cloud Spanner Sentinel Failure",
            "text": "Failed to query Cloud Monitoring for Spanner CPU metrics. Verify project ID and service account credentials.",
            "footer": "Kestra Workflow Failure Sentinel"
          }
        ]
      }

triggers:
  - id: periodic_schedule
    type: io.kestra.plugin.core.trigger.Schedule
    cron: "*/15 * * * *"
    disabled: true