2024 – Present · DevOps & Infrastructure · DevOps & Cloud · Self-hosted and maintained
ThinkCentre Proxmox Homelab & Three-VM K3s Platform
A compact self-hosted lab for Kubernetes, CI/CD, storage, ingress, and recovery practice
What this system is for
Hoan's current personal homelab is a small virtualized platform: one ThinkCentre host runs Proxmox, which provides three Ubuntu 26 virtual machines for the K3s cluster and self-hosted workloads. It is a hands-on environment for learning what stateful workloads, ingress, storage, delivery, and recovery feel like when the hardware, hypervisor, guest OS, and cluster all belong to one operator.
What I owned
Proxmox and VM topology, Ubuntu guest preparation, K3s operations, Jenkins delivery, registry and ingress configuration, Longhorn storage, troubleshooting, and release discipline.
Current homelab profile
This is the actual current lab topology, separate from the multi-node bare-metal clusters used by other systems.
- Host
- 1 ThinkCentre physical host
- Compute
- Intel Xeon E-2286M · 8 cores / 16 threads · 32 GB RAM
- Virtualization
- Proxmox VE
- Guest nodes
- 3 Ubuntu 26 virtual machines running the K3s nodes
Overview & Context
Hoan's current personal homelab is a small virtualized platform: one ThinkCentre host runs Proxmox, which provides three Ubuntu 26 virtual machines for the K3s cluster and self-hosted workloads. It is a hands-on environment for learning what stateful workloads, ingress, storage, delivery, and recovery feel like when the hardware, hypervisor, guest OS, and cluster all belong to one operator.
Problem & Constraints
A single physical host is a constrained but useful platform for learning the entire path from hardware and virtualization to guest operating systems, cluster scheduling, persistent storage, ingress, and application delivery.
System Architecture
┌─────────────────────────────────────────────────────────────┐
│ ThinkCentre · Intel Xeon E-2286M · 8C/16T · 32GB RAM │
│ Proxmox VE │
└──────────────┬────────────────┬────────────────────────────┘
│ │
┌────────▼───────┐ ┌──────▼────────┐ ┌────────▼────────┐
│ Ubuntu 26 VM 1 │ │ Ubuntu 26 VM 2│ │ Ubuntu 26 VM 3 │
│ K3s node │ │ K3s node │ │ K3s node │
└────────┬───────┘ └──────┬────────┘ └────────┬────────┘
└────────────────┴───────────────────┘
│
▼
┌─────────────────────────────────────────────────────────────┐
│ K3s · Traefik · Longhorn · Registry · Self-hosted Workloads │
└──────────────────────────────┬──────────────────────────────┘
▲
│ (Automated CI/CD Delivery)
Git Push ──► Jenkins Pipeline ──► Docker Image ──► K3s Rollout
Internet / Cloudflare Tunnel
│
▼
▼
Traefik / Ingress- Jenkins declarative pipelines with multi-stage Docker caching
- Proxmox VM topology with multi-node K3s deployments, services, and ingress
- Longhorn distributed block storage volumes with automated snapshots
- On-premises container registry, Traefik ingress, and health check probes
Engineering Deliverables
- Built multi-stage Docker build files with layer caching, reducing image footprint and build times significantly.
- Configured Jenkins declarative pipelines for automated linting, test execution, image packaging, version tagging, and automated Kubernetes manifests deployment.
- Built the current lab topology on one ThinkCentre host running Proxmox with three Ubuntu 26 VMs for the K3s nodes.
- Deployed Longhorn distributed block storage across the virtualized K3s nodes for persistent workloads and snapshots.
- Configured Traefik Ingress with TLS certificate automation, namespace isolation, ConfigMaps/Secrets, and health check probes (liveness/readiness).
- Handled cluster maintenance, CoreDNS troubleshooting, node draining, and rolling application releases.
Engineering Signals & Outcomes
- 1 ThinkCentre / Proxmox host with 3 Ubuntu 26 K3s VMs
- Longhorn distributed persistent storage across the virtualized cluster
- Automated declarative Jenkins pipelines for Docker image builds & K8s rollouts
- Rolling deployments and self-healing container infrastructure
Engineering Outcomes
- A compact Proxmox-based platform for testing delivery and stateful workload behavior on owned infrastructure.
- A realistic view of the full stack: physical host, hypervisor, Ubuntu guests, K3s, storage, and ingress.
- A clearer distinction between a successful manifest apply and a healthy service, while keeping the single-host failure boundary visible.
Engineering threads behind the project
Each thread links to notes focused on one concrete decision or engineering problem.
One host, three guests, one real operations boundary
Why the ThinkCentre → Proxmox → Ubuntu 26 VMs → K3s topology is useful for practicing cluster operations while making the single physical-host boundary explicit.
Operating the stateful parts, not only deploying YAML
The lessons from nodes, storage, ingress, registries, persistence, and recovery are more valuable than a list of Kubernetes objects.
Technologies & Tooling
ThinkCentre · Xeon E-2286M · Proxmox · Ubuntu 26 · Kubernetes · K3s · Longhorn · Docker · Jenkins · Traefik · Linux · Git · CI/CD