Homelab
Enterprise-style platform patterns, self-hosted
This is where I test architecture decisions, run reliability experiments, and validate operating patterns in a real multi-node environment before I trust them elsewhere.
Architecture Snapshot
Bare-metal + virtualization + Kubernetes with practical SRE/DevOps discipline.
Provisioning / IaC
OpenTofu (Terraform fork)
Orchestration
k3s Kubernetes
GitOps
FluxCD workflow model
Ingress / LB
Traefik + MetalLB
Storage
Longhorn + ZFS/NFS
Observability
Prometheus + Grafana + Uptime Kuma
Edge Security
Cloudflare Tunnel (Zero Trust)
Node Topology
- Titan (storage + compute profile)
- Sirius (worker profile)
- Vega (control-plane profile)
- Lyra (cluster node)
Multi-node design gives me realistic failure, placement, and service continuity scenarios to operate against.
How it’s run
- Git-driven infrastructure and service definitions
- Incremental change model with validation before promotion
- Monitoring-first mindset before scaling services
- Runbook-backed troubleshooting and recovery workflows
- Focus on repeatability over one-off manual fixes
Operational outcomes
- Faster troubleshooting due to centralized visibility and alert context
- Cleaner service exposure with ingress/load-balancer patterns
- Better resilience for critical app state via replicated storage
- Safer external access via zero-trust edge model
- Stronger confidence in automation and recovery procedures
Interview answer (ready to use)
I run a 4-node k3s homelab as a production-style platform. I use OpenTofu for infrastructure definitions, Proxmox for virtualization, Traefik + MetalLB for ingress/load balancing, Longhorn plus ZFS/NFS for tiered storage, and Prometheus/Grafana/Uptime Kuma for observability. Cloudflare Tunnel handles zero-trust external access. I use this setup to test deployment, reliability, and incident response patterns in a realistic environment.