Deployment¶

Docker Images¶

All application images are built for linux/amd64 and linux/arm64 (Raspberry Pi compatible).

Registries¶

Images are published to both Docker Hub and GitHub Container Registry on every release:

Service	Docker Hub	GHCR
Orders	`kkaarroollm/orders-service`	`ghcr.io/kkaarroollm/orders-service`
Delivery	`kkaarroollm/delivery-service`	`ghcr.io/kkaarroollm/delivery-service`
Notifications	`kkaarroollm/notifications-service`	`ghcr.io/kkaarroollm/notifications-service`
Simulator	`kkaarroollm/simulator-service`	`ghcr.io/kkaarroollm/simulator-service`
Frontend	`kkaarroollm/frontend-service`	`ghcr.io/kkaarroollm/frontend-service`

Tags follow semver: :0.1.0, :0.2.0, etc. The :latest tag always points to the most recent release.

CI/CD Pipeline¶

On PR / push to master: all 5 images are built for both architectures (no push) to catch build failures early
On GitHub release: images are built, tagged with the release version, and pushed to both registries

Docker Compose¶

The docker-compose.yaml defines the full development stack: application services, databases, monitoring, and reverse proxy.

Environment Files¶

Copy the defaults before first run:

cp envs/default.mongo_db.env envs/mongo_db.env
cp envs/default.redis.env envs/redis.env
cp envs/default.simulator.env envs/simulator.env
cp envs/default.mongo-keyfile envs/mongo-keyfile

Commands¶

# Start everything
docker compose up --build

# Restart after config changes
docker compose down -v && docker compose up -d

# View logs for a specific service
docker compose logs -f order-service

NGINX Proxy Routes¶

Path	Backend
`/`	Frontend (Vite dev server)
`/api/v1/orders`, `/api/v1/menu`	Order service
`/ws/v1/order-tracking`	Notifications service (WebSocket)
`/grafana/`	Grafana
`/prometheus/`	Prometheus (GET-only)

Kubernetes (Helm)¶

Chart Structure¶

charts/orders-project/          # Umbrella chart
  charts/
    orders/                     # Order service subchart
    delivery/                   # Delivery service subchart
    notifications/              # Notifications service subchart
    frontend/                   # Frontend subchart
    simulator/                  # Simulator subchart
    mongodb/                    # MongoDB StatefulSet
    redis/                      # Redis StatefulSet
  dashboards/                   # Grafana dashboard JSONs
  templates/
    configs.yaml                # Shared ConfigMaps
    secrets.yaml                # Shared Secrets
    ingress.yaml                # Ingress rules
    grafana-dashboards.yaml     # Dashboard ConfigMaps

External dependencies (from Chart.yaml):

ingress-nginx 4.12.1
kube-prometheus-stack 69.8.2
loki 6.28.0
promtail 6.16.6

Install¶

helm dependency update charts/orders-project
helm install orders charts/orders-project -n prod --create-namespace

Configuration¶

All values are in charts/orders-project/values.yaml:

global:
  mongo:
    username: root
    password: <change-me>
    host: mongo-svc
    database: food-delivery
  redis:
    password: <change-me>
    host: redis-svc

Init Jobs¶

On first install, three Jobs run automatically:

init-rs-job – initializes MongoDB replica set (rs0)
init-user-job – creates the MongoDB admin user
init-dummy-db-job – loads demo menu data

Ingress Routes¶

Path	Service
`/api/v1/orders`, `/api/v1/menu`	orders-svc:8003
`/ws/v1/order-tracking`	notifications-svc:8002
`/grafana`	grafana:80
`/`	frontend-svc:80

Monitoring in Kubernetes¶

ServiceMonitors are defined for orders, delivery, and notifications services. Grafana dashboards are provisioned via ConfigMaps with the grafana_dashboard: "1" label (auto-discovered by the sidecar).

Loki is configured as an additional Grafana datasource in values.yaml.

Raspberry Pi Deployment¶

All images are built for linux/arm64, so the full stack runs on a Raspberry Pi 4 or 5.

Resource Estimates¶

Stack	RAM	Disk
App only (5 services + MongoDB + Redis + NGINX)	~1 GB	~2 GB
App + monitoring (Prometheus, Grafana, Loki, Promtail)	~2-3 GB	~5-10 GB

The retention limits (Prometheus 3 days / 256 MB, Loki 72 hours) keep disk usage bounded over time.

Tips¶

A Raspberry Pi 4 with 4 GB RAM can run the full stack including monitoring
Use DEPLOY_ENV=prod for production-grade passwords
Consider reducing replicaCount to 1 in Helm values for single-node clusters
Cloudflare Tunnel works well for exposing the cluster without port forwarding