mo_br physical backup and restoration¶
Physical backups require consistent database state, metadata, and storage objects. Copying files from a running database directory is not a verified backup. Use a tool/workflow compatible with the database, deployment, and storage backend.
Availability and versions¶
The mo_br described here is an enterprise-service backup/recovery tool; see tool availability. Confirm program name/version, supported MO releases, storage format, full/incremental features, and restoration restrictions with the provider, then inspect the actual --help. Historical logs/package names are not evidence of an available, compatible v4.2.5 download.
This review did not obtain and execute the complete physical-backup toolchain. The requirements below are not a claim that v4.2.5 physical restoration passed. The runnable SQL snapshot experiment in recovery drills tests a separate recovery mechanism.
Full and incremental inventory¶
Type |
Retain |
Validate before restore |
|---|---|---|
Full |
Backup ID, recovery point, source version/configuration, object inventory, checksums, key references |
Independent media accessible, backup complete, target/format compatibility |
Incremental |
Base ID, ordered segment IDs/points, objects/dependencies |
Entire chain accessible; base/intermediate segments retained |
Save actual commands, completion status, errors, and duration, not only a backup ID. Compression, encryption, cross-platform, and cross-version support depend on the delivered tool guidance and tests.
Restore to an independent target¶
Select the recovery point and complete chain; check source/target MO/tool versions, FileService backend, node identity, and configuration.
Prepare an independent empty root/volume/bucket prefix and retain original production data. Stop target services/concurrent writes; keep backup media independent of primary storage.
Run commands validated for the current tool. Do not overwrite the old production directory or delete original
mo-datafirst.Verify target startup configuration, paths, access, and tool-required identities before startup. Restoring to a new path while
mo_ctl startstill launches the old path does not test the restored target.Check objects, counts, aggregates/watermarks, privileges, files, and background work; validate actual-client business behavior before deciding to switch.
Historical interface shape only; this is not a universal executable command and parameters require the actual tool’s help:
mo_br restore <backup_id>
--restore_dir filesystem
--restore_path <independent_empty_target_root>
Inappropriate repairs¶
Deleting original production
mo-data, PVCs, or shared buckets and assuming backups will recover them.Deleting Raft/WAL, checkpoint, or snapshot-referenced objects, or editing node identity arbitrarily.
Restoring concurrently into an active directory or sharing local state between two independent instances.
Declaring recovery from database names alone or treating backup frequency as measured RPO.
Record the complete chain, actual target, checks/business results, measured duration, and acceptable loss. See recovery acceptance for method selection, independent DR, and RPO/RTO.