Why we still write restore tests by hand
A backup you have never restored is a hypothesis, not a backup.
Server operations
Debian and Ubuntu fleets provisioned from code, hardened to a written baseline, patched in staged windows.
Infrastructure practice · Stockholm
We keep small production systems boring.
SCRF Systems looks after Linux servers for teams of five to fifty people. Patching, backups, monitoring and the unglamorous work that keeps a product online — handled on a fixed schedule, documented, and reproducible.
40+hosts under management
99.95%median availability, 2026
7 yrspractice running since 2019
< 15 minmedian first response
What we take care of
Four areas, one contract. No platform to migrate to, no proprietary agent on your machines.
Backup & recovery
Nightly incremental snapshots to a second provider, retention up to twelve months, restore drills every quarter.
Monitoring
Uptime, disk, certificate expiry and service checks, routed to email or chat with an on-call rota.
Reviews & migrations
Independent audit of what you run, a written risk list, and help moving between providers without downtime.
How an engagement runs
Most teams start with a two-week assessment and continue on a monthly retainer.
01 / ASSESS
Two weeks, read-only
We inventory hosts, services and data flows, then hand over a written report: what breaks first, what is untested, what costs more than it should.
02 / STABILISE
Fix the top risks
Backups that restore, alerts that fire, certificates that renew. Everything changes through review, with a rollback path.
03 / RUN
Monthly operations
Patching, capacity checks, incident response and a short monthly summary. You keep root access and every artefact.
Selected work
Names omitted by agreement; details available on a call.
Twelve hosts, one boring winter
Replaced hand-made deployments with a single Ansible baseline and moved backups off the same provider.
Restore time: 6 h → 25 min
Certificate renewals that stopped failing
Moved four domains to DNS-01, added expiry monitoring and a monthly check that actually runs.
Zero expired certificates in 14 months
Cutting a cloud bill without downtime
Moved static delivery to object storage and right-sized eight instances that had been sized for a launch that never came.
Monthly spend: −38%
Notes from the practice
Short write-ups, roughly once a month.
Certificate expiry is a people problem
Automation fails silently; the calendar reminder is what catches it.
Patching windows for teams without a night shift
How we schedule reboots around a business that runs 08:00–18:00.
Tell us what you run.
A short description is enough — number of hosts, what sits on them, what worries you.