merox Book a free callFree call

Your infrastructure, run like my own.

An ink drawing of me in glasses and a knit sweater

I’m Robert. I set up and look after the servers, network and backups of companies of 10 to 100 people. Everything documented, monitored and restorable.

Book a free call See the homelab

in infrastructure
8+ years
servers managed
250+
restore tests
Monthly

Based in Romania, working across the EU

Three things I fix first

Backups you can restore

A test restore every month, so you know your backups work before you need them.

At Netex: backups with Bareos, written to LTO tape.

My lab's backup plan

Alerts before the phone calls

Monitoring that reaches me when a service fails, a disk fills up, or a job stops running.

At Forvia: Prometheus and Grafana watching GPU clusters; at Netex, redundant switches, so one failure is not an outage.

How alerts reach me in my lab

Knowledge that stays with you

Infrastructure in code, clear documentation, and full access for your team.

At Netex and Forvia: Puppet, Ansible and SaltStack configuring hundreds of servers, the same way every time.

My lab, rebuilt from git
A production rack, drawn in three quarters: at the top two patch panels and two stacked switches with cable managers; under them a monitoring server, a configuration server, three hypervisors, the backup server and its LTO tape library, and a power strip near the foot.LTO600 mm800 mmSwitch ASwitch B, stackedMonitoringPuppet, AnsibleHypervisorHypervisorHypervisorBackups, BareosLTO tape library

How it goes

Three steps, each with its price agreed before it starts. Stop after any of them.

  1. Fixed price

    Audit

    I go through what you run and tell you what to fix first.

    • Servers, network and backups reviewed
    • The current risks, written down
    • A proposed solution and its cost

  2. Fixed price

    Build

    I set up what the audit proposed.

    • Linux servers, Proxmox and networking, declared in code and easy to rebuild
    • Monitoring, logs and backups in place
    • Documentation and full access, handed over

  3. Monthly fee

    Run

    I keep it running and show you how it went.

    • I know a service is down or a disk is filling up before you do
    • Updates, and a test restore every month
    • A monthly report

Start with an audit

And at home, a lab where I try it first

The rack beside my desk holds my own films, files, photos and backups. Nothing in it can be reached from the internet, and all of it can be rebuilt from a git repository.

Take the homelab tour

to restore the cluster from git
~35 min
at the wall, all of it
~65 W
in one small rack
7 devices
answering the internet
0 ports
Grafana, Power row: battery at 100%, 36 minutes of runtime, the estate drawing 74 W, and six hours of draw between 51 and 94 W
Power: what the rack draws, and how long the battery holds it.
Grafana, Internet row: 936 Mb/s down, 657 Mb/s up, 21 ms round trip, and six hours of flat throughput and latency history
The line: speed and latency, measured all day.

Captured from my Grafana on 4 October 2026.

Notes from the lab

What broke, and how I fixed it. Written down so it only has to be solved once.

  1. Spinning Down SAS DisksSAS drives ignore ATA standby and the usual tools quietly stop working on them. What actually parks them, and a script that keeps them parked. 12 minute read
  2. Offsite Backup to Oracle Cloud with resticAn encrypted offsite backup on an Oracle Cloud Always Free instance, no dependencies — in place of a HyperBackup setup that needed a working DSM to restore. 4 minute read
  3. One Identity Provider for EverythingAuthentik on Oracle Cloud in place of scattered logins: Google login everywhere, proxy auth for Guacamole, OAuth2 for Portainer, a K8s outpost for the cluster. 8 minute read

All notes

Still taking computers apart.

A drawing of a 2U server taken apart: the chassis with its drive cage, the lid lifted high above it, the fans, the processors under their heatsinks and the memory lifted out, two drives pulled from the front and the power supplies drawn out of the back.

It started with a Pentium 4 I kept breaking as a kid. Every job since has been the same habit on someone else’s hardware, with a budget and a change window.

Before it was a job it was a few years keeping Metin2 and CS 1.6 servers online for people who depended on them — the first time uptime was someone else’s problem too — and then a first sysadmin role in 2018 (full story).

The homelab is where I get to skip the change window: whatever I can’t try on production at work, I try there first, and most of what I write here starts as something that broke in it. The homelab tour shows what runs in it today.

Write to hello@merox.dev, or find me on GitHub and LinkedIn.

  1. 2022–today HPC Systems Administrator Forvia
    • 250+ servers; GPU clusters of NVIDIA A100 and H100 on InfiniBand
    • Ansible and SaltStack, from OS install to a node in service
  2. 2020–2022 Cyber Security Engineer Atos IT Solutions
    • Firewall rules on Palo Alto, Fortinet, Cisco and Juniper
    • Forescout policies for more than 4,000 users; Nessus assessments
  3. 2018–2020 Systems Administrator Netex Romania
    • Four full APC racks: racked, cabled and patched by hand
    • Puppet and Ansible across the fleet; Bareos backups to LTO tape

Download my CV (PDF)

Tell me what’s on your mind.

The first conversation is free and takes 45 minutes. Pick what sounds familiar and I’ll come prepared.

Pick any that apply

Opens your mail app with what you picked, addressed to hello@merox.dev. Nothing is sent from this page.

How I handle your email