Docker HEALTHCHECK Generator

The HEALTHCHECK instruction tells Docker how to test if a container is working. Docker runs the health check command at the specified interval. If the command exits with 0, the container is 'healthy'. If it exits with 1, it is 'unhealthy'. The health status is visible in docker inspect and docker ps, and used by orchestrators like Kubernetes and Docker Swarm for routing decisions.

start-period gives your container a grace period after it starts before health check failures count toward the retries threshold. During start-period, failures are recorded but don't contribute to the failure count. This prevents containers from being marked unhealthy during a slow initialization (e.g., database startup, JVM warm-up). Set it longer than your worst-case startup time.

curl is the most common: curl -f URL exits with non-zero on HTTP errors (4xx/5xx). wget -q --spider URL is a lighter alternative. If curl isn't in your image, install it, or use CMD-SHELL with a language-specific alternative. For minimal images, consider adding a tiny health check binary like healthcheck.

PostgreSQL: pg_isready -U postgres. MySQL/MariaDB: mysqladmin ping -h localhost. Redis: redis-cli ping. MongoDB: mongosh --eval 'db.runCommand({ping:1})'. These are purpose-built health check commands that don't require opening a full connection and are fast and low-overhead.

The Dockerfile HEALTHCHECK is baked into the image — it applies whenever the image is run. docker-compose healthcheck overrides or supplements the image's HEALTHCHECK for a specific service in a Compose setup. You can use depends_on with condition: service_healthy to wait for a service to be healthy before starting a dependent service.

Docker marks the container as 'unhealthy' in docker ps and docker inspect. The container keeps running — Docker does not automatically restart unhealthy containers. Orchestrators (Docker Swarm, Kubernetes) use health status to route traffic away from unhealthy containers and replace them. For Kubernetes, use readinessProbe and livenessProbe instead of Docker HEALTHCHECK.