Databáze řešení

By default, a crashed or stopped container simply stays stopped. Restart policies and healthchecks let Docker automatically recover failed containers and detect when an application is unhealthy even if the process technically hasn't exited.

Restart Policies

Set with --restart on the command line, or restart: in Compose.

Setting a Restart Policy

docker run -d --restart unless-stopped nginx

In Compose:

services:
  web:
    image: nginx
    restart: unless-stopped

Limiting Restart Attempts

docker run -d --restart on-failure:5 myapp

This gives up after 5 failed restart attempts, preventing an infinite crash loop from consuming resources.

Why Restart Policies Alone Aren't Enough

A container process can technically stay running while the application inside it is unresponsive (e.g. a web server accepting connections but timing out on every request). Docker's restart policy won't catch this — that's what healthchecks solve.

Adding a Healthcheck in Compose

services:
  web:
    image: myapp
    healthcheck:
      test: ["CMD", "curl", "-f", "http://localhost:3000/health"]
      interval: 30s
      timeout: 5s
      retries: 3
      start_period: 10s

start_period gives the app time to boot before failed checks start counting against it.

Adding a Healthcheck in a Dockerfile

HEALTHCHECK --interval=30s --timeout=5s --retries=3 \
  CMD curl -f http://localhost:3000/health || exit 1

Checking Container Health Status

docker ps

The STATUS column will show (healthy), (unhealthy), or (starting).

For detailed healthcheck history:

docker inspect --format='{{json .State.Health}}' CONTAINER_NAME

Using Health Status with depends_on

Ensure a dependent service only starts after the database is actually ready, not just running:

services:
  web:
    depends_on:
      db:
        condition: service_healthy
  db:
    image: mysql:8.0
    healthcheck:
      test: ["CMD", "mysqladmin", "ping", "-h", "localhost"]
      interval: 10s
      retries: 5

Common Errors

Container marked unhealthy despite working correctly — the healthcheck command may need adjusting (wrong port, endpoint, or tool not installed inside the container's image).

Infinite restart loop — use on-failure:N to cap attempts, and check docker logs for the underlying crash cause instead of relying solely on auto-restart.

Best Practices

• Use unless-stopped for essentially all production services
• Add healthchecks to any service another container depends on
• Set a reasonable start_period for apps with slow startup

FAQ

What happens if the Docker daemon restarts?
Containers with always or unless-stopped (and not manually stopped beforehand) will automatically restart when the daemon comes back up, including after a full VPS reboot if Docker itself is enabled at boot.

Related Articles

• Deploy Your First Docker Container
• Docker Compose Troubleshooting: Common Errors & Fixes
• Run MySQL/PostgreSQL/Redis in Docker Containers