- Single job instead of 3 (no artifact overhead)
- Build directly on k3s node (avoids image transfer)
- Parallel Docker builds with BuildKit
- Auto-create namespace if missing
- Automatic rollback on failure
- Health check via domain
- Add linux-musl-x64 runtime identifier for Alpine builds
- Install ICU libs for SQL Server globalization support
- Fix nginx upstream to use k8s service name (student-api)
- Sync package-lock.json with package.json
- Increase memory limit from 1Gi to 1.5Gi
- Increase MSSQL_MEMORY_LIMIT_MB from 768 to 1024
- Increase liveness probe initialDelaySeconds to 120
- Increase failureThreshold to 5 for more tolerance
- Increase readiness probe timeouts
Fixes DEF-001: SQL Server was crashing with Exit Code 137 (OOMKilled)
due to insufficient memory allocation for SQL Server 2022 Express.
Docker:
- Multi-stage Dockerfile for API (.NET 10)
- Multi-stage Dockerfile for frontend (Angular + Nginx)
- docker-compose.yml with resource optimization
- Nginx reverse proxy configuration
- Health checks for all services
Kubernetes (k3s):
- Namespace and ConfigMap
- SQL Server StatefulSet with PVC
- API Deployment with HPA
- Frontend Deployment
- Services and Ingress
- Network policies for security
- Secrets management
Resource optimization:
- SQL Server Express with 1GB RAM limit
- API with 512MB limit
- Frontend with 128MB limit
Andrés Eduardo García Márquez