Watchtower

A container-based solution for automating Docker container base image updates.

Quick start

Watchtower watches the Docker daemon it's pointed at, polls the registries of the running containers' images, and — when a digest changes — gracefully stops each stale container and recreates it with the same config (volumes, networks, env, command).

Pull from either registry; the images are identical:

docker pull openserbia/watchtower          # Docker Hub
docker pull ghcr.io/openserbia/watchtower  # GitHub Container Registry

Run it

docker rundocker-compose

docker run --detach \
    --name watchtower \
    --restart unless-stopped \
    --volume /var/run/docker.sock:/var/run/docker.sock \
    openserbia/watchtower \
    --interval 60 --cleanup

services:
  watchtower:
    image: openserbia/watchtower:latest
    container_name: watchtower
    restart: unless-stopped
    volumes:
      - /var/run/docker.sock:/var/run/docker.sock
    command:
      - --interval=60
      - --cleanup
      - --label-enable

That's enough to start: Watchtower polls every 60 s and deletes old images after a successful update.

Scoping what it updates

By default, every container on the host is a candidate. To opt in only a chosen set, pass --label-enable and label the target containers:

docker run -d --label com.centurylinklabs.watchtower.enable=true my-app:latest

Omit the label to keep a container untouched (databases, stateful services, anything with a manual release process).

What's next

• Arguments — every flag and WATCHTOWER_* env var
• Container selection — include/exclude by label, scope, or name
• Notifications — Shoutrrr, email, Slack, Teams, Gotify
• Private registries — credentials via config.json
• HTTP API mode — /v1/update, /v1/metrics, /v1/audit endpoints
• Metrics — Prometheus exposition format + shipped Grafana dashboard
• Lifecycle hooks — pre/post scripts inside the target container
• Why this fork? — what openserbia/watchtower adds over upstream