[Borgbackup] Consistent unexpectedly large deduplicated size

[Thomas Waldmann]

> One of them has a consistent unexpectedly large deduplicated size, much larger than the sum size of the new files being added.

Besides file content data, borg also backs up file metadata.

Usually this isn't much, but in some cases it could contribute 
significantly to the overall size:

- large xattrs or ACLS changing

- files which you think are still "the same" (because that have same 
content data) have different metadata now and produce a different borg 
metadata stream (not well-deduplicating against the metadata streams of 
past backups). the usual reasons for this are: archiving atime, 
different mount point / different base dir, chown/chmod -R, different 
mtime ("touch").

- file discovery: dir tree traversal order, unstable inode numbers

- file size vs. disk usage: sparse files can be very large size and use 
little disk space.

you have excluded atimes from the backup archive, so it is not that in 
your specific case.

> Are there any ideas about how to find out?

You can check "borg debug ..." whether it has enough to analyze this.

-- 

GPG Fingerprint: 6D5B EF9A DD20 7580 5747  B70F 9F88 FB52 FAF7 B393