One deploy panel for 13 production servers

The situation

A client’s products were spread over 13 servers at four hosting providers, each set up by hand: nginx here, Caddy there, apps started with pm2 or systemd. Deploying meant SSH and memory.

What I did

• Read before installing. Dokploy’s stock installer would have stopped on busy ports, left the servers’ cluster, and let its proxy grab ports 80 and 443 at boot — taking every site down. Instead I attached each server alongside its existing web server, with the panel’s proxy bound to a local port and guards so it can never take the public ports.
• A fleet CLI. 24 commands for snapshots, before/after diffs, external port probes, attaching servers and moving apps.
• Safe app moves. Each app is deployed side by side, tested against the running copy for every domain, and only then switched over; the old copy stays for seven days and rollback takes seconds. Secrets are compared by hash and never printed.
• Real client IPs. I fixed IP handling through the extra proxy hop so rate limits and logs stay correct.

Result

All 13 servers sit in one panel, and 11 production apps — SaaS backends, Next.js sites and APIs — moved over in two days, with responses checked identical before and after. Along the way I found and logged an expired certificate and a system updater that had been stuck for 165 days.