Re: [Bacula-users] Problems with incremental backups

[Sebastian H. <ha...@em...>]

Hello,

Phil Stracchino wrote:
> Sebastian Haas wrote:
> 
>>I've did a incremental backup last night and I notice that the backup
>>was bigger than I did assumed.
>>
>>I did then an "bls" to show which files bacula backed up, and the most
>>of the files bacula backed up shouldn't changed to the last backup. By
>>the way, I did a full back yesterday.
>>
>>But "bls" showed me that the date of files is 2029-10-31. I checked the
>>date of this files on the filesystem and the dates are in the range of
>>1998-2003. Some files has been really changed and therefore backed up
>>right, but the most files bacula backed up hasn't changed.
> 
> 
> This is more complex, and may be harder to diagnose.  I believe we've
> had one similar case before in which a locale problem was causing Bacula
> to get invalid times back, but I don't recall the exact details.
> 
> Can you provide logs of the backups?  You might need to restart Bacula
> with debugging enabled and run an incremental backup to generate a log.
Currently not, but I gave it tomorrow.

>>As the jobs started I've noticed the following message from the director:
>>29-Sep 00:52 fileserver01-fd: DIR and FD clocks differ by -730 seconds,
>>FD automatically adjusting.
>>
>>The director started the job at 1:05am while the FD's clock was 00:52.
>>But I would assume that this is not a problem.
> 
> 
> These two items are related.  You have a clock drift problem on your
> network.  You might want to consider establishing a timeserver and
> syncing all your internal machines to it, even if you don't actually
> sync your timeserver to an external authoritative timeserver.
Okay, fixed!

Let us assume the problem is fixed due to the synchronized clocks, will
an incremental backup (tonight) save files that it saved last night if
they changed or will it not?

What I try to say is, how bacula finds out which files needs to be
backup incremental? On a per timestamp basis or on a content basis (e.g.
MD5, SHA1, ...)?

Thanks in advance.

-- 
Mit freundlichen Gruessen/Best Regards,
Sebastian Haas