Monitoring¶

Orion includes built-in Prometheus metrics collection and Grafana dashboards.

Stack¶

graph LR
    S["Services\n/metrics"] --> P["Prometheus\n:9090"]
    P --> G["Grafana\n:3003"]
    G --> U["Dashboard\nAlerts"]

Component	Port	Purpose
Prometheus	9090	Metrics collection and storage
Grafana	3003	Visualization and alerting

Prometheus Configuration¶

The Prometheus config is at deploy/prometheus.yml:

global:
  scrape_interval: 15s
  evaluation_interval: 15s
  external_labels:
    project: orion

scrape_configs:
  - job_name: gateway
    static_configs:
      - targets: ["gateway:8000"]
    metrics_path: /metrics

  - job_name: scout
    static_configs:
      - targets: ["scout:8001"]

  - job_name: director
    static_configs:
      - targets: ["director:8002"]

  - job_name: media
    static_configs:
      - targets: ["media:8003"]

  - job_name: editor
    static_configs:
      - targets: ["editor:8004"]

  - job_name: pulse
    static_configs:
      - targets: ["pulse:8005"]

  - job_name: milvus
    static_configs:
      - targets: ["milvus:9091"]

Grafana Dashboards¶

Pre-provisioned dashboards are in deploy/grafana/provisioning/dashboards/json/:

Dashboard	File	Description
Orion Overview	`orion-overview.json`	Service health, request rates, latencies
Provider Health	`provider-health.json`	AI provider status and costs
GPU Resources	`gpu-resources.json`	GPU utilization (when using --profile gpu)

Datasource¶

Grafana is auto-provisioned with a Prometheus datasource:

datasources:
  - name: Prometheus
    type: prometheus
    access: proxy
    url: http://prometheus:9090
    isDefault: true

For detailed configuration, see the Monitoring section.