More recoveries begin with "We had backups" than with any other sentence. Not "We had no backups" (the honest doomed, who at least have an understanding of where they stand), but the past tense of people who discover, in the middle of a disaster, that their backup system was merely decorative. In twenty years of having this conversation with clients, we've categorised the failures into five species. Hopefully, you'll recognise your own before it counts.
Mirrored drives. Stale scheduled jobs. Sync traps. Single points of failure. Unrestorable backups, and the one test that finds them all.
1. The mirror. A RAID system. A synced NAS. A duplicate drive mirroring another: an availability tool masquerading as a backup solution. Mirrors will accurately replicate (at the speed of light) the same corruption and deletions as the drive they mirror. Ransomware and all other forms of corruption can be replicated with equal velocity. 2. The stale job. A backup schedule that has been silent since March. Destination drive has reached capacity. Credential to access destination drive has expired. Update has disabled scheduled task. Discovered in November. 3. The sync trap. Folders in cloud syncing services, where files are deleted locally on a user’s device and then are instantly deleted on their cloud “backup”. If the service allows versioning, this may provide some assistance in retrieving a file that was deleted, but only if you notice before those versions expire. This typically requires users to set up the service in advance so that versioning is enabled. 4. The single point. A backup drive located next to a workstation, subject to the same fires, floods, surges and burglaries as the workstation itself. Or a NAS device that is the only copy, where there exists no redundant storage of the same data. 5. The unrestorable. The backup job that was never test-restored. The encryption key used to create the backup is now lost. The backup has contained corrupted data for months. The backup job did not include the one critical folder, due to the fact that it was added to the source after the creation of the backup job.
Every species above will die at the hands of a single predator: the test restore. Not checking whether or not the process worked, but restoring an actual file, from an actual backup to a completely different machine, and then opening it. That proves the copy exists, is up-to-date, can be read, is complete and is accessible without the source system: five confirmations in five minutes. Do this on a quarterly basis; document when you do it; assign someone to perform it. The 3-2-1 model (three copies, two types of media, one located outside your site) is the design for backups, but testing the ability to restore a backup is the audit, and a backup that has never been tested is, statistically speaking, a wish.
Found this page because your backup has just failed you? Two quick truths. First: the failed original plus the failed backup are both recovery points. A stale backup with a recovered delta from the failed disk can often be used to fully rebuild the data, so save both and advise the bench what date range each represents. Second: the computer which first identifies a backup failure will likely be asked to "try" things for the next sixty minutes, and those sixty minutes of attempts at rebuilding your disaster will do far more damage than the actual disaster itself. Shut it off. Inventory all available copies. Let the free diagnostic find the union of what remains.
Not certain of your species? A test restore will let you know that in five minutes. When it's already told you the hard way, the bench will be able to inform you as to how much this will cost. The diagnostic is completely complimentary, your quote will be fixed and in writing, and there's honest advice for absolutely no charge at all on 0131 202 0491.