Central vs local backup scope — farm edge stack
Parent package: Backup and disaster recovery package — smart farm stack. DR decisions: Disaster recovery decision rules — farm edge stack.
Definitions
| Term |
Meaning here |
| Central |
Primary backup landing zone you treat as authoritative for recovery (home-base NAS, object store, cloud bucket)—often SITE_HOME in the two-site model. |
| Local |
On-gateway or on-field storage used for queueing, short retention, or survival when WAN is down—not a substitute for off-device central copies unless policy explicitly accepts single-site risk. |
What should be backed up centrally (typical)
| Asset class |
Why central |
| farmOS / PostgreSQL logical dumps |
Farm records SoR—survive gateway swap or DB host loss |
| restic repository or object-store target for logical exports |
Versioned recovery; encryption at rest per tool |
| Longhorn backup target (S3/NFS) when used |
Off-node PVC history; pairs with k3s story |
| k3s etcd snapshots (embedded etcd ) |
Control plane DR—store off the server disk (S3-compatible per k3s docs ) |
| Rancher backup artifacts (if Rancher installed ) |
Management UI + CR bundle per Rancher Backups docs |
| Helm values / GitOps repo (non-secret) |
Reproducible redeploy |
What can stay local (with eyes open)
| Asset class |
Local role |
Risk if local-only |
| Edge gateway SQLite / buffer DBs |
Survive brief outages |
Field device loss = buffer loss if never replicated central |
| MQTT broker spool (if any ) |
Transient ingress |
Usually not SoR—confirm against SoR page |
| restic repo on USB disk at farm |
When WAN unreliable |
Site fire / theft—still need off-site policy for true DR |
| Mesh / LoRa configs |
Fast field recovery |
Document export to central Git or encrypted archive |
Two-site pattern (East Tennessee scenario)