Re: backup on volume much larger than fileset

Hi Stefan,

Am 18.03.2026 um 09:59 schrieb Stefan G. Weichinger via Bacula-users:
...
> With this Fileset:
> 
> Fileset {
>    Name = "ivy-projects"
>    Include {
>      File = "/home/sgw/projects"
>      Options {
>        Signature = "Sha1"
>      }
>    }
>    Exclude {
>      File = "/home/sgw/projects/packer/my-work/packer_sgw_builds/ 
> packer_output"
>      File = "/home/sgw/projects/docker/qemu-docker-1/qemu_storage/ 
> debian/*.iso"
>      File = "/home/sgw/projects/packer/builds/cache"
>    }
> }

one of the things I would do is to verify that the excludes do work as 
expected. As far as I know, globbing is not supposed to work, for 
example. As I'm typically using Options with matching and exclude=yes to 
exclude, so I don't have anything to compare against.

> While I am quite sure that the JobDef is not the issue
 > ... > Now, what's the problem?
> 
> As you can see above, I try to backup the directory "ivy:/home/sgw/ 
> projects" minus some excluded files etc
> 
> # du -sh /home/sgw/projects
> 29G    /home/sgw/projects
> 
> But the resulting backup on bacula is ~ 164 GB in size!
> 
> Why that?
> 
> I browsed the directory for links etc ... nothing special.

That inspired me to do a quick and dirty shell command (one line):

C=0 S=0 find . -type f -printf '%s\n' | while read f ; do (( S = S + $f 
)) ; (( C++ )); if ! (( C % 50 )) ; then echo $C: $S ; fi  ; done

> In the details of the Fileset:
> 
> OneFs: true
> Recurse: True
> HardLinks: True

Of course my counting above is ignoring all those finer points, but with 
the discrepancy you observe, that shouldn't matter much :-)

> If I disable Hardlinks, the Estimate looks the same as before.
> 
> I browsed the Files in the Job to spot errors, with ~400.000 files 
> that's a bit tough.
> 
> That directory is located in a btrfs-pool, might that play a role here?

Definitely. I've never really used btrfs, but the commands

btrfs filesystem du

and

btrfs filesystem df

probably exist for a rason :-)

COW-filesystems, snapshots, sophisticated space and volume management in 
file systems make the traditional approaches of checking file system 
occupancy a bit too tricky.

Let us know what you find, I'm definitely curious ;-)

Cheers,

Arno

> For comparison I created a second task for the parallel folder 
> "projects_legacy". This one works great, same settings etc -> correct 
> size on the volume(s).
> 
> Any hints here? Thanks a lot, this would save me time and space on storage!
> 
> 
> 
> _______________________________________________
> Bacula-users mailing list
> [email protected]
> https://lists.sourceforge.net/lists/listinfo/bacula-users

-- 
Arno Lehmann

IT-Service Lehmann
Sandstr. 6, 49080 Osnabrück



_______________________________________________
Bacula-users mailing list
[email protected]
https://lists.sourceforge.net/lists/listinfo/bacula-users