[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [Bacula-devel] Automatic labeling (bug 1014)

Hash: SHA1

Thanks.  I understand the value of it being unique, but to me when
comparing the impacts of the two bugs I assumed that the autolabel
algorithm changing was the larger problem.

I'll check out the Python method, thanks.


Kern Sibbald wrote:
> On Monday 28 January 2008 19.45:55 Luke Reeves wrote:
>> Greetings all.  I've been working at tracking down what I thought was a
>> bug in Bacula.  Automatic volume labels prior to Bacula 2.2.5 were based
>> on the number of volumes in the pool where the new media is created and
>> in that version a change was made so that the incremental component of
>> the label was based on the maximum media ID across all volumes.  The
>> original behavior is restored by slightly modifying the SQL in newvol.c:
>> SELECT MAX(MediaId) FROM Media,Pool WHERE Pool.PoolId=%s
>> To:
>> SELECT COALESCE(MAX(MediaId), 0) FROM Media WHERE PoolId=%s
>> So just as I was about to mail that, I found bug 1014
>> (http://bugs.bacula.org/view.php?id=1014) which describes this as
>> changed behavior and the documentation was updated to reflect that.  I
>> can't tell from the bug if this is intentional or not, but to me it
>> seems very much like a regression.  
> No, it is not a regression.  If you delete a volume, the previous algorithm 
> completely fails.
>> Consider a site with multiple pools, 
>> one for each type of backup.  You might have the numbers in your labels
>> scattered across all of your pools, so instead of "FULL0001, FULL0002,
>> INC0001, INC0002" you would have "FULL0001, INC0002, FULL0004, INC0004".
>>  This is also a problem since labels for the tapes are usually ordered
>> in bulk (like say FULL0001 to FULL0120) and skipped numbers will just
>> lead to confusion.
> Well Automatic volume names is just that, Bacula *automatically* generates the 
> volume names and all that really counts for it to work is that they are 
> unique.  Automatic labeling was never guaranteed (even with the old 
> algorithm) to produce unique sequential label names ...
> If you want volume names to follow a specific pattern, then you need to either 
> pre-label them manually (or using barcodes), or use something like Bacula 
> variables or Python scripting to ensure that they are sequential.
>> I found a reference on newvol.c to a bug 921, but I'd think that the fix
>> for that causes a larger bug.
> Regards,
> Kern
Version: GnuPG v1.4.5 (Cygwin)
Comment: Using GnuPG with Mozilla - http://enigmail.mozdev.org


This SF.net email is sponsored by: Microsoft
Defy all challenges. Microsoft(R) Visual Studio 2008.
Bacula-devel mailing list

This mailing list archive is a service of Copilotco.