How remla2025-team10/operation works
Project type and purpose
• This is a full-stack web application for sentiment analysis of restaurant reviews.
• It comprises two microservices (app-service for the frontend+backend, model-service for ML inference) and extensive infrastructure code (Docker Compose, a Vagrant/Ansible-provisioned Kubernetes cluster, and a Helm chart with monitoring).
File structure
Top-level directories and files:
• Vagrantfile + ansible/: Infrastructure as code for spinning up a multi-node Kubernetes cluster (control plane + workers).
• docker-compose.yml + app-ingress/: Local Docker Compose deployment of app-service and model-service (with an NGINX ingress).
• deployment.yaml, finalization.yml, kubernetes-dashboard/: Raw Kubernetes manifests and playbooks for cluster finalization and dashboard setup.
• model-stack/: A Helm chart defining Kubernetes Deployments, Services, Ingress, ServiceMonitor, AlertManager rules, etc., for production-style deployment.
• model.json: The trained ML model artifact.
README for additional insights
• Two deployment modes: local Docker Compose vs. Vagrant-provisioned K8s vs. Minikube + Helm.
• The Kubernetes cluster uses Flannel (network), MetalLB (LoadBalancer), NGINX Ingress, the Dashboard, Prometheus + AlertManager + Grafana.
• App-service listens on port 8080 and proxies sentiment requests to model-service on port 3000.
Essential architectural elements to capture
• Infrastructure provisioning (Vagrant/Ansible) and Kubernetes cluster layout
• Service mesh boundaries (app vs. model service)
• External access path (DNS → MetalLB → Ingress → app)
• Monitoring & alerting pipeline
• Dependency on Helm for deployment templating