I set up a Arch Linux box with an external drive(Ext4) and netatalk 3, to be my backup server.
My conf file contains just this:
[TimeMachine]
path = /media/backup
time machine = yes
It has been running for a few weeks now, and I already had trouble twice.
First I got "resource temporarily unavailable", which was resolved by rebooting everything.
Now, time machine failed to verify my backup and prompted me to start a new one. I'm not sure how to debug this, but an unreliable backup is as good as no backup at all.
I've experienced this, too, on an OpenIndiana (oi_151a5) box running netatalk 3.0 as part of napp-it. All filesystems are ZFS, including the TM volume. My backups go bad about every 10-14 days, but only on the machine that uses encrypted backups and Mountain Lion. I have two other Macs (one Lion, one ML, neither using encrypted backups) backing up to this same sharepoint and they haven't had this problem.
I have the same problem. Once a month (when Lion fscks the 'sparse volume' for the time machine backup), my backup becomes 'unreliable'. Happened less often before I upgraded to netatalk 3.0.1.
So the question remains if TM corrupts the disk, netatalk does, or fsck doesn't work correctly.
I found this blog post about how to fix broken sparsebundles on disks connected to an airport extreme. http://www.garth.org/archives/2010,07,16,124,fixing-time-machine-sparsebundle-network-backup-errors.html
To me this suggests TM might be at fault, as the airport extreme obviously doesn't use netatalk.
I also seem to recall that no sparebundle is needed on HFS+ volumes, but this might have been with local disks only.
Tentatively closing this. The reports are against ancient versions of netatalk. I've been running hourly backups for two days with 3.1.15 pre-release code, and a Ventura Mac, so far without incident. I will keep this running for at least a month to see how reliable it is.