Re: [BackupPC-users] against which basis rsync do compare files before transmit the changed part?

[Holger P. <wb...@pa...>]

Hi,

Matthias Meyer wrote on 2009-01-28 08:30:54 +0100 [[BackupPC-users] against which basis rsync do compare files before transmit the changed part?]:
> I do not fully understand how rsync works within backuppc.
> In the documentation for BackupPC_dump I found:
>  As BackupPC_tarExtract extracts the files from smbclient or tar, or as
>  rsync runs, it checks each file in the backup to see if it is identical to
>  an existing file from any previous backup of any PC.

this refers to the pooling mechanism. Any file identical to "an existing file
from any previous backup of any PC" is never written to disk, not even
temporarily or partly. If you think about it, describing tar/smb/rsync in one
sentence *can* only refer to pooling, because you don't get any savings
transport-wise for identical content with tar/smb.

> Did that mean that during the rsync-process each sourcefile will be compared
> against all files in pool and cpool and if a file is found theire than only
> the changed part of this file will be transmitted?

No. How would you do that? "Changed part" means you are talking about a file
not already in the pool. Relative to which pool file would you expect the
differences to be minimal? :-)
Even finding an *identical* file *in the pool* is not possible for reasons
outlined many times on the list (let's not start that discussion again).

> Or did rsync compare against all files in all backups of the appropriate
> host

No.

> or only against all files in the last full backup of the appropriate
> host?

No.

Against the *same file* (meaning same path) in the reference backup (last
backup of next lower level for incremental, previous backup of any level for
full). Note that if you have an incremental as reference, the file may, in
fact, be "inherited" from an earlier backup, if it was not changed in the
reference backup itself. This is only a technical detail. The comparison is
done against the state the file was in at the time the reference backup was
made. If you are thinking about the bandwidth used (or the time transferring
this bandwidth takes), that is the relevant point. For all files, the delta
of changes since the reference backup needs to be moved over the line. You
don't get any savings for the same delta having being moved before - whether
it was for another file in the same backup or for a backup of a different
host.

Regards,
Holger