Backup and recovery
The operator view of what CALIBER state must be preserved, what can be rebuilt, and where disaster recovery stops and asset-specific rollback begins.
Use this page when you need the operator view of what CALIBER state must be preserved, what can be rebuilt, and where the disaster-recovery boundary stops and asset-specific rollback begins.
At a glance
| Surface | Preserve it? | Why it matters |
|---|---|---|
| Metadata database | yes | control-plane records, governed state, releases, jobs, and references |
| Object store data | yes | uploads, previews, extracted artifacts, and log sink material |
| Workflow storage | yes | run workspaces and file-backed workflow evidence |
| Generated docs/UI assets | rebuildable | these come from repository sources and build outputs |
1. Back up the state that cannot be re-derived safely
The safest minimum backup set is:
- metadata database
- object-store content
- workflow storage
Without those together, a restore may recover only part of the product state.
2. Distinguish disaster recovery from feature rollback
Recovery of the deployment itself is not the same as rolling back a governed asset release. Use the disaster-recovery plan to restore the platform, then use asset-specific rollback or release controls where needed.
3. Verify restores, not just backups
The useful operator question is not “did a backup job run?” but “can this deployment restore to a usable, trustworthy state?”
After a restore rehearsal, verify:
- readiness
- storage access
- release and review history visibility