Playground
A private server platform built one layer at a time. Each layer has to be tested, written up, and proven to survive a reboot before the next one starts.
The idea
A backup you have never restored is not a backup.
A lab with acceptance criteria
Playground is a private infrastructure environment where I can practise platform engineering on something real. Services go in a few at a time. Each one has prerequisites, resource limits, network boundaries, and checks that have to pass before it counts as done.
Backups get tested, not assumed
Backup coverage gets inspected, recovery copies get validated, restore paths get run, and critical layers get checked again after a reboot. When something fails, the method and the notes change to match.
Written for whoever comes next
Each finished service leaves a record: what it is for, what it depends on, how it recovers, and the evidence used to accept it. If I am the only person who can run it, it is not done.
What this work demonstrates
Evidence from the work.
-
A backup that wasn't covering everything.
A coverage audit found a service whose own state lived outside the backed-up path. The gap was closed and the audit became a scheduled check with tiered exit codes.
-
A script that reported success while failing.
A pull job exited zero regardless of transfer result. Caught, fixed, and turned into an acceptance condition for every subsequent job.
-
Restores get run on a schedule.
Recovery copies are validated and restore paths run regularly, well before an incident.
-
Reboot survival is an acceptance criterion.
A service isn't accepted until critical layers are re-verified after a cold boot.
-
Backups pull, they don't push.
The backup host reaches in over a restricted, command-forced account. A compromised source cannot reach the backups.
Public boundary
The diagrams show the layers and the method. Addresses, topology, credentials, access rules, and backup destinations stay private.