My UrBackup server is running on Linux Mint 19.2 kernel version 4.15.0-54 (/root and /home on 120 GB SSD, backup on mirrored btrfs 3 TB HDDs) backing up 5 Windows 10 clients. I not backing up system partitions on clients, just data partitions. Most of the files being backed up are in the 500 KB to 25 MB range.
During initial file and image backup server received around 1.05 TB of data which translated into roughly the same amount of used space on btrfs volume (info obtained with df command). Is this normal? There is significant number of identical files on different clients. Or do they count as different due to different file ownership and permissions?
During second file backup server received around nothing from 2 clients (this is ok I guess as there has been no user activity on them) an around 10 GB from remaining 3 clients. Now this is odd for two reasons. First, there was minimal user activity on them so no more than few files where changed/created. Second, this translated into additional 10 GB used on btrfs volume.
No, identical files should get linked. You can verify that by looking at the files with filefrag -v (physical offset is identical), or by looking at a server debug log file (it says Linked file … to file …). If you run internet client it also logs about the amount of files it did not load.
Could be okay, could be not. Often there is a surprising amount of churn on the clients. There are also some surprising btrfs snapshot space trade-offs. Make sure you mount btrfs with compress-force. I’d recommend using kernel 4.19.x with compress-force=zstd.
Now I’m really puzzled. For the files I have checked filefrag -v confirms that deduplication is working, yet when I sum amount of data being backed up from clients and size of backup on server they are pretty much identical.
Further backups worked fine, expected amount of data backed up matched amount of data being backed up.
Most of the data I’m backing up are .jpeg, .doc, .docx and .dwg files, would compression really help?