Prometheus metrics

Kanchi ships with a Prometheus endpoint so you can scrape task and worker telemetry straight from the backend — no extra plugins required.

Endpoint

The FastAPI backend exposes metrics at /metrics (plain text, Prometheus format).

curl http://localhost:8765/metrics

Scrape configuration

Add Kanchi as a scrape target in your Prometheus config. Example for a single instance:

scrape_configs:
  - job_name: kanchi
    metrics_path: /metrics
    scrape_interval: 15s
    static_configs:
      - targets: ["kanchi:8765"]  # replace with your host/port

If you run multiple Kanchi instances behind a load balancer, point Prometheus at each instance directly or use service discovery.

Metrics emitted

kanchi_task_events_total (counter)
Count of task events observed from the Celery broker. Labels: task_name, event_type, worker.

kanchi_task_queue_wait_seconds (gauge)
Time a task spent queued at a worker before execution began. Labels: task_name, worker.

kanchi_worker_prefetch_count (gauge)
Number of tasks currently prefetched (reserved) by a worker. Labels: task_name, worker.

kanchi_task_execution_duration_seconds (histogram)
Duration of task execution. Labels: task_name, worker.

kanchi_worker_status (gauge)
Worker availability flag (1 = online, 0 = offline). Labels: worker.

kanchi_worker_active_tasks (gauge)
Count of tasks currently being processed by a worker. Labels: worker.

Usage notes

• Metrics are derived from Celery task and worker events; ensure workers emit events (start Celery with -E or set worker_send_task_events=True).
• Task and worker names flow into labels. If you have highly dynamic names, consider Prometheus relabeling to keep cardinality manageable.
• The endpoint is unauthenticated by default. Place it behind your usual network controls or a reverse proxy if you need to restrict access.