Zero-Downtime Deployments with Docker Compose

The goal

A restart that takes your site offline for a few seconds is a deployment that failed quietly. With a reverse proxy in front and a health check after it, a Compose stack can be updated while the old container keeps serving traffic.

Health checks come first

Compose lets you declare a healthcheck for every service.

• The proxy should only pass traffic to containers reporting "healthy".
• The application's health endpoint must not just return 200; it should verify the database connection behind it.
• Give the check a generous start period on the first boot.

The update strategy

When you run `docker compose up -d`, Compose recreates changed containers one by one. Pair that with a running proxy and you get a rolling update for free.

• Run the migrations and other long tasks as a separate one-shot service before the web container restarts.
• Let the new container reach healthy before the old one is removed.
• Watch the logs during the switch instead of walking away.

What this blog's deploy actually does

The deploy script for this site pulls the new images, runs a migration guard, starts the new containers and then verifies a health URL returns 200 before reporting success. That verification step is what turns a deploy from a wish into a fact.

The takeaway

Zero downtime is a composition of small habits: a real health endpoint, a proxy that respects it, and a deploy script that will not celebrate early. Automate the checks and your teammates will thank you.