Jobs disappear from restore list

[Harald_Bratko]

We are using Veritas BackupExec for Windows Server on an Overland Neo 2000 Library with two Quantum SDLT600 Tape drives. Our scheme performs incremental backups of our machines each day together with a weekly full backup. All jobs work fine and are finished successfully, however, as I watched the restore list, I noticed, that successful jobs disappeared from that list.
An example: Machine A is backuped each day. When I examined the restore list on December 28, the restore list includes backups from December 24, 25, 26, 27, 28. On December 29, the backup from December 24 disappeared from that list; on December 30, I couldn't find the backups from December 24, 25 and 26, however, the backups from December 27, 28, 29 and 30 where present.
It seems that catalog entries were deleted or overwritten. The whole restore list staring from June 2004 periodically lacks of up to 15 sequencing jobs - but there is no pattern. I should mention that this concerns several tapes and that we do not truncate or delete our catalogs.
Does anyone have the same problems or an idea that could help solving it?

[Gilly]

Harald,

Are you 100% sure that your not Truncating/Deleting the Catalogs?
Might be worth checking again.
In Ver9.1 click Tools>Options
Under the settings Heading click on Catalog to view Truncation/Deletion settings.

Also check Database Management under the same Settings header.

To be honest I don't know of any other way your Catalogs could be overwritten!

Gilly

[Harald_Bratko]

Gilly,

thanks for your quick response, but I'm sure that I don't truncate/delete catalogs - in fact, this was the first issue I examined when I first noticed this problem; I now checked it once again - both options are disabled.
The point is, that not only old or unused backups dissappear - it seems to be rather accidental. So, for example, from one of our servers today I would be able to restore data from January 6 and 5, and December 28, 23, 20, 19, 18, 17, 14, 1. The same fragmentary (but other dates) list for November, October ...
I think, database settings are also correct; I allow deleting expired data (default settings), but, from my point of view, this only concerns protocols and announcements and should not affect the restore list. And even if - the most recent announcement that is allowed to be deleted is 7 days old - I am not able to restore data that was backuped three days ago.

Harald

[Gilly]

It might be worth Cataloging a tape from one of the dates that is not visible and see does this overwrite another catalog.

In any case this is a very strange problem.
Have Veritas provided you with any help on the matter?

[Harald_Bratko]

I have already tried this which leads to a "Data on medium is inconsistent" error. I think, this can mean everything - maybe that the tape is damaged, but than we have a drop out of nearly all of our tapes.
I'm in Email contact with (German) Veritas support but until today they were also not able to solve the problem - I think, since this obviously is not a known problem, it's very difficult for them to find the reason - it could be anything - software bug, hardware error, tape error - but, I must say, they are really exerted.
And - you are right - it's really a strange problem.

[Gilly]

Might be nothing:

http://news.support.veritas.com/dnewsweb.exe?cmd=article&group=veritas.backupexec.windowsnt.english&item=76494&utag=

Found this while searching for Data on Media is inconsistent error.
One guy said
"It was all down the fact that my raid controller and Dat unit were on the same SCSI"

He put in a new card and seperated them and so far its working for him.

Gilly

[Harald_Bratko]

Thanks for your help, Gilly, but our raid is attached to another machine and the backup server's SCSI adapter is exclusively used by the backup devices. But following your link I now know, that I'm not the only one dealing with this problem: "SOME OF MY CATALOGS ARE DISAPPEARING" ...

[Harald_Bratko]

We now have found the reason of our problems: if a job that backs up more than one drives (or servers) cannot connect to the first drive but successfully to another one in the line, BackupExec resets the set counter for storing the backuped data. Thus BE starts to overwrite all stored data (or at least the database entries used for identifying the restore selection). Thus from the point the counter is reset, the corresponding tape is no longer usable for restoring the data backuped so far.
As an example consider a notebook user that connects his laptop to the network just at that time the backup job starts to work - to late to connect to the first drive, but accurately timed for backing up the second drive - this will reset the set counter and make the tape unusable for restoring data stored so far.
We perform the incremental backup of our notebooks in one job at the same time each day - everytime the first notebook is not connected to the network and BE can successfully connect to another one, BE resets the set counter.
As a workaround, We now backup a small file directly from the backup server at the beginning of each job - this guarantees, that BE can connect successfully to the first drive and we have no problems anymore. We do this for each job because we want to avoid that a reboot of any server at the wrong time will corrupt the backup data of a whole tape.
From our point of view this behaviour is definitely a bug in the Veritas BE software.