Bartu Ogur

// homelab dossier

The homelab & the stack

ONLINE

A cluster of Intel nodes at home — 4 Proxmox hosts and a 40 TB Unraid server running a 6-node Kubernetes lab (5 workers + a control plane, one worker per machine) across ~68 vCPU and ~190 GB of its ~380 GB of RAM, on OpenEBS-LVM with the Unraid for bulk NAS storage (NVMe-cached). It sits behind OPNsense and Traefik with a full Prometheus / Grafana / Loki stack, and it is where I prototype the platform patterns I run in production and self-host my own world — GitLab, dashboards, media and a heap of apps. A small data centre, only slightly exaggerated; upgrading it is a happy addiction.

hypervisor
Proxmox
kubernetes
6 nodes
compute
~68 vCPU
memory
~380 GB
gpu
RTX 3090 + 3060
nas
Unraid · 40 TB
power
UPS

// the metal

// service map

// homelab — service map all traffic routes through the platform

Homelab service map. Traffic enters from the internet through an OPNsense firewall/VPN and a Traefik ingress into a 6-node Kubernetes platform. Running on it: dev/CI services (GitLab, Harbor, ArgoCD, n8n); AI and agents (LiteLLM gateway, Enki — an enterprise AI framework, AutoDev, Sagent, MCP servers, and image generation on ComfyUI with FLUX.1, Fooocus and DDColor); apps/knowledge tools (BookStack, Vikunja, Linkwarden, SearXNG, Stirling-PDF, Homarr, Moodle); web and business apps (Ghost CMS, FreeScout help desk, Mautic, Vaultwarden); and a self-hosted, automated media library (Jellyfin, Navidrome music streaming, a PeerTube video instance, the Seer request portal, a Mullvad privacy browser, and a library-automation stack); and persistent multiplayer game servers run for friends (Terraria, Core Keeper, Don't Starve Together, and Ark: Survival Ascended). They share a foundation of observability (Prometheus, Thanos, Grafana, Loki), storage (OpenEBS-LVM and a 40 TB Unraid NAS), data services (PostgreSQL, Redis, ClickHouse, Airbyte and a custom ETLProcesser), and UPS-backed power.

// toolset · by layer

cloud · finops

  • AWS (Solutions Architect Pro)
  • Amazon EKS
  • VPC / Route 53 / ACM / WAF
  • S3 / EFS / RDS / OpenSearch
  • KMS envelope encryption
  • Savings Plans & RIs
  • Right-sizing

orchestration

  • Kubernetes (kubeadm, CKA in progress)
  • Calico CNI
  • Helm
  • Terraform
  • Docker
  • Rook-Ceph / CephFS
  • Traefik & NGINX
  • MetalLB
  • cert-manager
  • Velero

ai · ml-serving

  • vLLM
  • Multi-GPU inference pools
  • LiteLLM routing & keys
  • MLOps
  • Agentic platform engineering
  • Per-user workspace pods
  • Qdrant (RAG)
  • pgvector (RAG / memory)
  • LangChain / MCP / A2A
  • Ollama / llama.cpp
  • GGUF (Unsloth)

data · streaming

  • ClickHouse
  • PostgreSQL / TimescaleDB
  • Redis (Sentinel)
  • Kafka
  • RabbitMQ
  • Python ETL
  • Airbyte / Meltano
  • Apache Superset

observability

  • Prometheus (agent) + Thanos
  • Grafana
  • Loki
  • Alertmanager
  • GPU / model telemetry
  • Uptime Kuma

security

  • RBAC & NetworkPolicy
  • AWS KMS / key management
  • OWASP ZAP / Burp Suite
  • WireGuard VPN
  • OPNsense / pfSense
  • Multi-tenant isolation
  • ISO 27001 / 27002

net · storage

  • Proxmox / ESXi / KVM
  • OpenEBS LVM
  • Ceph / ZFS / NFS
  • Unraid
  • TrueNAS
  • OPNsense
  • Technitium DNS
  • Restic / rsync
  • Route 53 / Cloudflare

languages · eng

  • Python
  • Go
  • Bash
  • SQL (query & index tuning)
  • TypeScript
  • GitLab CI/CD & GitOps
  • k8s-native service design
  • AI-augmented dev