Backup storage¶
Register a named destination for installation backups, InfiniDysk 2 major-migration archives or standalone and scheduled symlink snapshots. Open Settings → Backups and restore for the central controls. These appear when the backend advertises the corresponding storage capabilities. Existing backup paths remain the default until you explicitly select a registered destination.
Storage at a glance¶
| Storage | Suitable for |
|---|---|
| Local, private storage | Active databases, controller recovery state, inventory/restore scratch |
| Registered local or already-mounted NAS directory | Closed installation/migration archives and symlink snapshots |
| Host directory not mounted into DUMB | Unavailable until exposed through the deployment |
Registering a destination does not make it the default or move an existing backup. Choose it explicitly in the operation or save it in the relevant schedule.
Register an existing destination¶
Make the directory visible to DUMB before registering or selecting it:
flowchart TD
accTitle: Register storage before selecting it for backups
accDescr: Expose an existing directory to the container, pass the destination preview, register it, and explicitly select it for an operation or schedule.
Host["Prepare local or mounted NAS storage"] --> Visible["Expose its existing directory inside DUMB"]
Visible --> Enter["Enter a unique destination ID and container-visible path"]
Enter --> Preview["Preview destination"]
Preview --> Passed{"Checks passed?"}
Passed -->|No| Fix["Resolve the path, mount or permission blocker"]
Fix --> Preview
Passed -->|Yes| Register["Register destination"]
Register --> Select["Select it for the operation or save the schedule"]
Select --> Verify["Review the chosen destination before running"]
The directory must already exist inside the DUMB container. A host-only path is insufficient. Mount NFS/SMB storage on the host and expose it to the container through your normal deployment configuration before registration. DUMB does not mount network shares or request their credentials.
- Expand Register a backup destination in Settings → Backups and restore, the InfiniDysk migration panel or the Symlink operations panel.
- Enter a unique identifier, an optional display name and an existing container-visible path. For a trusted destination with fixed/shared permissions, review Allow shared archive permissions, described below. It starts unchecked.
- Select Preview destination, review its checks, then Register destination. A failed check stays visible so you can correct it and retry.
- Select the destination for the current operation. The separate Global default controls show the saved destination; use Set global default or Clear global default deliberately. Registration alone changes neither that preference nor any service's backup selection.
Use global default explicitly adopts that saved preference. Clearing the global default blocks future operations using that choice until you select a new default or an explicit destination. An existing custom backup path remains active when the legacy/local choice is selected. Saving a symlink schedule applies its selected destination to that schedule; changing a draft selection does not save it. Use a new identifier when moving an existing registered root to another directory or filesystem.
Manage saved destinations¶
Backends supporting destination management show Registered destinations with each destination's path, default badge, last check and any blockers. Not checked, Available at last check and Unavailable at last check describe the latest explicit check; listing or refreshing the catalog does not probe the NAS. Select Check availability to check the current mount. An operation error stays with its destination until you address it.
- Edit changes the display name, enabled state and, when supported, shared archive-permission consent. The identifier and path stay fixed. Select Save changes or Cancel; editing does not move files or change another operation's destination/default. A name change can be saved without reaching the storage.
- Remove registration opens a review. Confirm remove registration removes the saved choice, while archive files, the owned marker and local ownership/catalog receipts remain. It does not delete backups.
- Disabling or removing a destination is blocked while a saved default, configured consumer, active operation or retained recovery still needs it. Review the listed reason: change the relevant selection or finish/recover the operation first. Do not purge recovery records to bypass a blocker.
To use a removed destination's retained catalog again with the same DUMB configuration, register the same identifier and original storage path. The original identifier stays associated with that owned storage; use a new identifier for a different root. If the old controller state/catalog is unavailable, installation restore also supports an existing mounted archive file.
Mount checks after a container update¶
Linux can assign a different device number to the same mounted share after a container is recreated. Current backends can recognize that change when the registered storage coordinates, directory identity/permissions and owned marker still match. Existing archive ownership and references stay intact; you do not need to remove the destination for this device-number change alone.
A mount change while the same controller is running remains blocked until reviewed. After Check availability, an eligible destination offers Review mount change. Compare the registered and observed storage, confirm that it is the same archive storage containing the retained backups, then select Confirm controller mount check. A review expires after five minutes; changed storage or settings require a fresh review. A disabled destination can be reviewed without enabling it, then enabled separately.
This refreshes the controller's check for the same owned storage. It cannot adopt a different share, directory or marker, move backups, or bypass permission checks. Missing/replaced storage remains blocked. Check the host mount and container-visible path before retrying; DUMB never writes into the local directory underneath an absent registered mount.
Cold archives and local working storage¶
Local directories and already-mounted NFS/SMB directories can hold closed archive files. DUMB requires restrictive POSIX permissions by default. A trusted mount or deliberately shared local directory can instead use Allow shared archive permissions when the backend supports it. This applies only to the selected destination, including its managed markers, temporary exports and completed artifacts; existing profiles keep the restrictive default.
Some SMB mounts report fixed directory/file modes even when the server applies separate access controls. Observed modes and registration warnings are not proof of NAS ACL privacy. See the Linux CIFS documentation. Enabling the option acknowledges that you trust the mount/share's access controls. DUMB does not change the existing mount's ownership or permissions, verify server ACLs, or encrypt these exports.
Archives and partial exports can contain configuration credentials, identity/signing keys and database data. Restrict access to the share and its clients before registering it, protect downloaded copies, and keep an independent backup with a tested restore procedure. DUMB's local catalog and controller recovery state still require private local permissions.
Active SQLite databases, migration control state, the inventory manifest and import/database-restore scratch stay on verified local storage. SQLite WAL requires local shared-memory and locking behavior and does not support network filesystems. See the SQLite WAL documentation. Selecting NAS archive storage therefore does not remove local working-space requirements. Review each filesystem's capacity before starting.
The registered-storage consumers are the tools listed above. Existing Arr PostgreSQL migration backups, native PostgreSQL recovery snapshots, application-managed backups and runtime-install rollback transactions keep their existing locations and settings.
For a managed installation archive, scheduling and fresh-container recovery,
see Installation backup and restore. Configuration
.bak copies, service-reset configuration backups and migration recovery bundles
remain separate tools. Protect host deployment files, external systems and media
through their independent backup/restore plans.
Verification and recovery¶
After a backup completes, check that it appears in the relevant archive or snapshot picker. Open its receipt and confirm that verification passed. If the destination is unavailable, restore the original mount and use Check availability before retrying.
Technical details: artifact identity, verification, and catalog publication
DUMB records the selected root's identity and writes only its own managed artifacts. Exports use newly created temporary files, complete-byte checksum/readback verification and atomic publication. The selected permission policy does not bypass file-type, owned-identity or checksum checks. The manifest browser lists registered artifacts rather than browsing arbitrary NAS files. Opening or downloading a managed artifact verifies its content again; the list's checksum records the completed export verification.
For registered destinations, a missing or replaced mount blocks use instead of writing into the local directory underneath an absent NAS mount. Legacy custom paths retain their existing behavior until you select registered storage. After an interruption, preserve the job, partial-export evidence and original source, then restore storage availability and review the offered recovery action. Network availability and server durability still depend on your deployment.
Managed backups use opaque references. Select them through the picker or use the authenticated download API; the reference is not a website URL or a filesystem path. By default, InfiniDysk creates its local import candidate from the original source; automated major rollback starts the retained originals until that rollback is explicitly retired.
The migration panel's Migration backup location selects the recovery archive location. Selecting a share alone does not move the original v1 data. On supporting backends, Reuse v1 directory avoids a second local source copy; Move legacy blobs to migration backup location also copies and verifies legacy blob input on that storage before releasing its owned local files. Both require a registered destination on a different filesystem and a reviewed plan.
The verified archive is the v1 recovery source in both modes. Expanded legacy input is a separate working copy, so the share needs capacity for both when offload is selected. Active SQLite, DUMB control state and PostgreSQL verification scratch remain local. Rollback restores v1 from the archive and can remove the owned v2 candidate to make room first. Keep the destination mounted through the operation and any recovery; a missing expanded input does not replace or invalidate the separate verified archive. DUMB attempts to remove only its recorded unchanged working copy after accepted import or completed v1 restoration. A retained staging-cleanup warning does not authorize deleting changed files. See migration storage choices.
Scheduled symlink retention removes only that process's scheduled managed snapshots. Manual snapshots are retained separately from automatic pruning. A symlink snapshot records link paths and targets; protect the referenced media and application databases through their normal backup plans.
See InfiniDysk 2 installation and upgrade, Symlink operations and the Process Management API.