Application recovery drills¶

The application recovery module creates manual backup and isolated restore-drill services from application-owned hooks. A recipe defines a consistent export, backup transport, recovery selection, restore operation and semantic check. No production application is stopped until its backup service is explicitly run.

Read the module and operational guide before declaring a recipe. The generated options describe the configuration interface. Keep credentials in protected runtime files and provision real backup destinations before enabling a recipe.

The disposable tests restore committed SQLite data through Restic after deleting the original, and check that VM disk, definition, firmware and virtual TPM fixture state belong to the same backup. The NixOS test checks service shutdown/restart and prevents the drill from writing to live directories. These results cover the fixtures; actual application recovery and a real guest boot need separate drills.

Check health without a desktop session¶

The desktop collector also checks enabled network, DNS, login, SSH and Nix services. Additional command and freshness probes can cover application receipts. A successful process-activity probe establishes service availability only; it does not prove a working application workflow.

An independent system timer delivers warnings and recovery transitions through an operator-provided command. Failed or unconfigured delivery retains a pending record. Configure a real channel, run its test-delivery command, and confirm receipt from the destination before relying on alerts while logged out.

Use validation records to keep fixture, native build, runtime and real recovery evidence distinct.