Automatic rollback trigger criteria on Fedora Atomic variants

Bootloader-Based Automatic Rollback

Fedora Atomic variants support automatic rollback on boot failure through systemd-boot or GRUB with boot counting. When configured, a deployment that fails to boot can automatically fall back to the previous deployment without manual intervention.

Failure Detection Uncertainty

The current implementation relies solely on boot success to determine deployment validity. This approach does not detect application-level failures, service crashes, or schema-migration issues that manifest after the system has successfully booted. The bootloader cannot distinguish between a system that boots cleanly but fails to operate correctly versus one that boots normally.

Deployment Retention Limits

rpm-ostree retains only a limited number of deployments (typically current and previous), meaning older states are garbage-collected. This shallow rollback depth combined with uncertain failure detection creates ambiguity about when and how automatic rollback should trigger.

• What specific boot or service failure conditions should trigger automatic rollback on Fedora Atomic systems?
• How can administrators configure boot counting and fallback behavior to handle partial failures that occur after successful boot?