restic scan is returning unexpectedly large result

[jkahrs]

When backing up a webserver with many html-root dirs I get a bigger scan result than expected

Output of restic version

restic 0.2.0 (v0.2.0-181-g24385ff-dirty)
compiled at 2016-09-07 14:22:18 with go1.7

Expected behavior

restic scans through the directories and gets a combined filezize close to df.

Output of df:
~# df -h
Dateisystem Größe Benutzt Verf. Verw% Eingehängt auf
rootfs 443G 324G 97G 78% /
/dev/root 443G 324G 97G 78% /
devtmpfs 4,9G 0 4,9G 0% /dev
tmpfs 1001M 164K 1001M 1% /run
tmpfs 5,0M 0 5,0M 0% /run/lock
tmpfs 2,0G 0 2,0G 0% /run/shm

Number of webspaces:
~# ls -l /var/www/ | wc -l
2308

Actual behavior

restic scan finishes at about 3x the used space on / after about 50mins and then begins backing up data

Steps to reproduce the behavior

start a backup on a "larger" web server

restic -r sftp://backup@[....]:2244/restic backup /var/www

[wscott]

Perhaps your data has a large number of hard links and restic doesn't bother tracking those since the deduplication will eliminate the extra data anyway?

Just guessing at possible reasons. I don't know how hard links are actually handled.

[jkahrs]

@wscott thank you for the tip. I rechecked the structrue within the directorys and actually found some hard links.
As far as i got it from https://github.com/restic/restic/blob/master/src/restic/archiver/archiver.go hardlinks wouldn't be handled while scanning.