Restoring a completed backup version results in an error

I have created backup of the foundationDB successfully, if I check the status of the backup below is what I get

The previous backup on tag `default' at file:///data/fdbbackup/backup-2019-12-16-12-43-45.243778 completed at version 21655918762.
BackupUID: 4b6d2cf2e97bd4a673b32a008ddaa17d
BackupURL: file:///data/fdbbackup/backup-2019-12-16-12-43-45.243778

But When I try to restore the same backup using below command fdbrestore start --dest_cluster_file /var/dynamic-conf/fdb.cluster -r file:///data/fdbbackup I get below error

No restore target version given, will use maximum restorable version from backup description.
The specified backup is not restorable to any version.
ERROR: Restore error
Fatal Error: Restore error

I am not sure why the backup is not restorable to any version.

Are you using -r file:///data/fdbbackup/backup-2019-12-16-12-43-45.243778 or just -r file:///data/fdbbackup ? The first points to this specific backup, the second points to a folder containing one or more backups.

I think I see the source of confusion: When you specify a file://URL to use a local directory destination, a subdirectory is created using the current timestamp and that subdirectory is used as the actual backup destination. This is legacy behavior that I thought was removed a while ago. The idea was to make it easy for an automated tool to take repeat backups using the same destination directory and have the fdbbackup tooling create subdirectories automatically. In practice, it has just been confusing to users.

1 Like

I opened a GitHub issue about removing this confusing behavior:

Hi @SteavedHams
That makes sense, I was actually just giving the parent directory that is file:///data/fdbbackup, after your suggestion I used file:///data/fdbbackup/backup-2019-12-16-12-43-45.243778 and things seem to work as expected, the data was restored.

Just wanted to highlight that the operator that we have for foundationDB doesnt support backup and restore mechanism by default, I was having this conversation with @john_brownlee as well. I have raised an issue to track that.