Stop All MLOps Services

Primary Action

Run the stop script, which handles the full shutdown sequence (Docker Compose first, then Minikube) in the correct order:

./scripts/stop-mlops-services.sh

Monitor the output for any [ERROR] lines. The script prints a final status check confirming everything is down.

Script Flags

• •--keep-minikube — Stop Docker Compose services but leave Minikube running (useful if you plan to restart compose services soon without the overhead of recreating the K8s cluster)
• •--remove-volumes — Remove Docker volumes during shutdown (DESTRUCTIVE: deletes Postgres data, MLflow experiments, Airflow history). The script prompts for confirmation before proceeding.

Verification

After the script completes, confirm everything is down:

minikube status 2>&1 || true
docker ps --format 'table {{.Names}}\t{{.Status}}'

• •Minikube should report "not found" or show no profile
• •No mlops-related containers should be running

Troubleshooting

Script hangs during docker compose down

A container may be stuck. Cancel the script (Ctrl+C) and force-remove the hung container:

docker rm -f <hung_container_name>

Then rerun ./scripts/stop-mlops-services.sh.

Minikube delete fails

If minikube delete errors out:

minikube delete --all --purge

If that also fails, force-remove the minikube Docker container:

docker rm -f minikube

Some containers still running after script completes

Force-stop and remove them:

docker rm -f $(docker ps -q --filter "name=mlops-services")

Partial Shutdown

To stop just one service without bringing down the whole stack:

cd /home/ubuntu/health-predict/mlops-services
docker compose --env-file /home/ubuntu/health-predict/.env stop <service>

To restart it later:

docker compose --env-file /home/ubuntu/health-predict/.env up -d --no-deps <service>

Always use --no-deps when targeting a single service to avoid the docker-compose v1 ContainerConfig bug.