Cluster mode

In cluster mode, the main image serves the web interface and the API, and workers run the checks. The workers share a job queue stored in PostgreSQL, so no extra service is needed.

Set up the application

Run the uptime-kotlin image with SCHEDULER_STRATEGY: none. It no longer runs checks itself, and it can be replicated behind a load balancer.

Add workers

Add one service per worker with the worker image:

uptime-kotlin-worker:
  image: ghcr.io/theomeunier/uptime-kotlin/worker:latest
  container_name: uptime_kotlin_worker
  restart: unless-stopped
  environment:
    TZ: Europe/Paris
    SCHEDULER_STRATEGY: database
    SCHEDULER_WORKER_NAME: worker-primary
    SCHEDULER_WORKER_CONCURRENCY: '4'
    QUARKUS_DATASOURCE_USERNAME: uptime-kotlin
    QUARKUS_DATASOURCE_PASSWORD: change-me
    QUARKUS_DATASOURCE_JDBC_URL: jdbc:postgresql://postgres:5432/uptime-kotlin
    ENCRYPTION_MASTER_KEY: change-me-32-characters-minimum-0
  depends_on:
    - uptime-kotlin

SCHEDULER_WORKER_NAME must be unique per worker. Checks left by a stopped worker are picked up by the others after 30 seconds.

The workers must use the same ENCRYPTION_MASTER_KEY as the application, and the same maintenance and notification settings (see Configuration).