Gene Heskett wrote at 08:53 -0400 on Apr  9, 2009:

 > Uptime is about 5 days now, but this may be the beginning of the
 > end.  Something made it think all data was new from the looks of
 > this.  This was the first run of 20090323, 20090321 works fine.
 > Another device mapper screwup?  It was updated by yum yesterday.

It probably was a device numbering change - if you are in the habit of
rebooting after a yum update.  It's probably worthwhile noting the
disk device numbering before / after a reboot.  Then if it changes,
you can repair the gnutar-list files to point at the new device to
avoid the "everything has changed" incremental dump (or 'amadmin
force' a level 0, but that has problems with scheduling balance not to
mention potentially running out of room depending on tape / vtape /
holding disk size).

I think there was a script floating around to change the dev # in the
gnutar listed incremental files - possibly Dustin's.