Production Kubernetes Platform on Bare Metal

YUKLID needed a production-grade platform, but cloud costs were prohibitive and I wanted to understand how everything actually works under the hood—not just follow tutorials.

Set up a complete Kubernetes cluster on physical servers. Implemented service mesh (Istio), distributed storage (Ceph), CI/CD pipeline (Jenkins), monitoring (Prometheus + Grafana), and security layer with automatic SSL and centralized auth (Keycloak). Not a managed service—built every layer from networking to observability.

Learned that building infrastructure isn't about memorizing kubectl commands—it's about understanding how distributed systems fail. When MetalLB wouldn't assign IPs, I had to debug networking at the packet level. When Ceph storage crashed, I learned about consensus algorithms the hard way. Best lesson: always monitor what you can't see.