Moving the server while it runs

Request: „Move the document archive to the new server — without stopping it.“

1. Advise on the hardware purchase, install virtualisation (Proxmox), run a burn-in test.
2. Write the migration as a runbook and move services in waves.
3. Set up the backup chain: seven steps every night, one encrypted copy off-site, restores proven regularly.
4. After two power cuts in one day, bring the uninterruptible power supply into service the same day.
5. Two alarm layers: one says what is broken; an outside watchdog says that something is broken.

Data sat in three clouds and on old drives; backup meant hoping.

The purchase decision, the go for every migration wave, pulling the plug.

The document archive once stood still for 30 hours 34 minutes — since then an outside watchdog raises the alarm when the server goes quiet.

A live migration succeeds with a runbook, waves and proven restores, not with one big weekend.

Instead of three clouds, there is a server in the basement that the AI helped me choose, set up and migrate twice. 35 services run there on eight virtual guests: document archive, Git, cloud storage, smart home, dashboard. Moving the document archive to the new machine cost about two minutes of downtime. Every night seven backup steps run, one encrypted copy sits off-site, and restores are proven regularly rather than assumed. When the power failed twice in one day, the uninterruptible power supply was in service that same evening.