2026 Production Voice Systems · intermediate
A Home Server Became Production the Day I Needed a Recovery Time
The difference between a lab box and production infrastructure is not the hardware; it is whether failure has a documented recovery path.
I stopped calling the server a lab machine once services on it had users, state and dependencies I cared about restoring.
The first step was inventory. Containers are easy to recreate only if I know which compose file, environment, volumes and external secrets they depend on. Database backups need restore tests. An encrypted archive that has never been opened on another machine is only a hopeful file.
I started separating configuration in Git from runtime state and secrets. Recovery documentation records the order: restore host prerequisites, deploy configuration, restore state, verify health, then expose traffic. That order matters because a service that starts is not necessarily a service that recovered correctly.
The useful metric became recovery confidence rather than backup count. If I cannot explain how to rebuild the box without the original disk, the backup story is incomplete.