Homelab

The Lab: Why a Home Server Is the Best Place to Learn Infrastructure for Free

May 22, 2025

homelab

proxmox

kubernetes

devops

self-hosting

learning

The lab

I posted a simple list on LinkedIn: the services running in the lab—my home server. No pitch deck, no architecture diagram in the thread. Just names.

That list is more than software. It is a private datacenter where I break things safely, learn for free, and ship experiments that would cost real money in the cloud.

If you are a developer curious about DevOps, Kubernetes, networking, or automation, a home server is one of the highest-leverage investments you can make—not for mining or nostalgia, but for education you own.

What “the lab” is

The lab is a homelab: infrastructure I run at home on hardware I control. The hypervisor is Proxmox; everything else is VMs or containers on top. I can snapshot before a bad upgrade, clone a service to test a migration, and reboot the router without filing a support ticket.

It is the opposite of a disposable cloud trial that expires in 30 days. The lab stays up, accumulates scars, and teaches you how production actually feels—messy dependencies, DNS that lied, disks that filled at 2 a.m.

The stack (and what each piece teaches)

flowchart TB
  subgraph edge["Edge & access"]
    AG[AdGuard Home]
    NPM[Nginx Proxy Manager]
  end
  subgraph platform["Platform"]
    PX[Proxmox]
    P[Portainer]
    K3s[K3s]
  end
  subgraph cicd["CI / automation"]
    GL[GitLab]
    N8N[n8n]
  end
  subgraph data["Data"]
    PG[(PostgreSQL)]
    RD[(Redis)]
    IFX[(InfluxDB)]
  end
  subgraph observe["Observe & enjoy"]
    GF[Grafana]
    HA[Home Assistant]
    JF[Jellyfin]
  end
  subgraph ai["AI gateway"]
    LL[LiteLLM]
  end
  PX --> K3s
  PX --> GL
  AG --> NPM
  NPM --> K3s
  NPM --> GL
  K3s --> PG
  K3s --> RD
  GL --> K3s
  IFX --> GF
  LL --> PG

Service	Role in the lab	What you learn by running it
Proxmox	Hypervisor, VMs, backups	Virtualization, storage, snapshots, VLANs
AdGuard Home	DNS filtering on the LAN	DNS, DHCP integration, privacy, blocklists
Nginx Proxy Manager	Reverse proxy + TLS	HTTPS, host routing, certificates
Portainer	Docker UI (where used)	Containers without jumping straight to K8s
GitLab	Git + CI/CD	Pipelines, runners, registries (ARM runner on a Pi included)
K3s	Lightweight Kubernetes	Deployments, services, ingress, Helm—without a managed control plane bill
n8n	Workflow automation	Integrations, webhooks, event-driven glue
LiteLLM	LLM API gateway	Routing models, keys, quotas—AI infra on your terms
PostgreSQL	Relational data	Backups, migrations, connection pooling in real apps
Redis	Cache / queues	Ephemeral state, pub/sub, performance patterns
InfluxDB + Grafana	Metrics & dashboards	Time series, alerting, SRE instincts
Home Assistant	Smart home hub	MQTT, automations, local-first IoT
Jellyfin	Media server	Storage, transcoding, LAN streaming—the “fun” workload that keeps you honest about bandwidth

None of this requires an enterprise license. Almost everything on the list is open source or free for personal use. You pay for power, hardware, and your time—not per-vCPU billing anxiety.

Benefits you do not get from tutorials alone

1. End-to-end ownership
In the cloud, networking, DNS, and certificates are often pre-chewed. At home, you connect AdGuard → NPM → the service. When it breaks, you learn the full path a packet takes.

2. Safe failure
Snapshot the VM. Break the cluster. Restore. No invoice, no account manager. That loop builds confidence faster than any certification slide deck.

3. Real constraints
Limited RAM teaches right-sizing. A single disk teaches backups. A residential uplink teaches why you cache and why edge matters.

4. Portfolio you can demo
“I run GitLab + K3s + observability at home” is a credible interview story—especially if you can explain trade-offs, not just name tools.

5. Freedom to experiment
Spin up LiteLLM, wire n8n to webhooks, test a GitLab pipeline on native ARM—projects that would be awkward or expensive as always-on cloud sandboxes.

Learning “for free” (what that actually means)

Free does not mean zero cost. You might spend on a used mini PC, a Raspberry Pi, or an old workstation. Ongoing cost is mostly electricity and time.

What you avoid:

Monthly fees for five managed services you only needed to study
Surprises when a forgotten load balancer runs all month
Vendor-specific abstractions that hide how things work

What you gain:

Skills that transfer to any employer’s stack
Reusable automation (Ansible, Terraform, GitOps) tested on your metal
Services you actually use daily—media, home automation, private CI—not just homework clusters

The lab is a tuition-free campus where the curriculum is whatever you need for work next quarter.

How to start without my exact list

You do not need fourteen services on day one. A sane progression:

One machine + Proxmox (or bare Linux if you prefer simplicity)
AdGuard or similar DNS control—immediate win for the whole household
Nginx Proxy Manager—one HTTPS URL to one app; understand reverse proxies
One workload you care about—GitLab or K3s or Jellyfin, not all at once
Metrics when pain appears—InfluxDB + Grafana when “why is it slow?” becomes real

Add K3s when containers outgrow docker-compose. Add GitLab when you want pipelines. Add LiteLLM when you are tired of juggling API keys in every script.

My list is what my lab grew into—not a shopping list for week one.

Why K3s fits home labs

Lightweight Kubernetes distros like K3s matter because full Kubernetes on a NUC is homework; K3s is homework you can finish. You still learn Deployments, Services, Ingress, and upgrades—but without the control-plane overhead of a “real” cluster you cannot afford to leave running.

That is why stacks like mine show up more often in homelabs: the same concepts as production, fraction of the resources. (And yes—managed platforms exist if you outgrow DIY; the lab teaches you when you actually need them.)

The lab in one sentence

Proxmox underneath, sensible edge (DNS + proxy), then the services you would otherwise rent—CI, orchestration, data, observability, automation, media, and home control—on hardware you can reboot without a change advisory.

It is not minimal. It is deliberate: a playground where every line on the list is a skill I wanted and could learn without asking permission.

If you have been thinking about a home server, start small, snapshot often, and let the stack grow with your curiosity. The cloud will still be there when you need scale—but you will understand what you are scaling.

Published originally on LinkedIn.