I have just encountered a serious problem with Netbackup that they couldn't
fix so I thought I would ask about it. I would be curious to see how other
backup systems handle this type of problem. If this is the wrong group to
post to please forgive me but I can't find any  *.backup groups.

Background: 
 We have several Netware, WinNT, Solaris and HPUX systems backing up to 3
slaves and one master (total of 4 backups servers). There are 2 DLT 35/70
drives attached to each system.

Problem: 
 One of our Netware servers' backups didn't register in the database. When
we went to do a restore the image was not listed as one we could choose
from. The other servers were fine. The backup had run without any errors and
had backed up 9 DLTs worth of files. The server in question had 110
gigabytes and 1 million files on it. It was spread across 2 DLT tapes.

 When we did the phase 1 import there were no problems. When we did the
phase 2 import (where it reads the tape and re-creates the database file) it
would run for 15 hours (about 8 hours a tape) and then fail with a
"Unexpected EOF" media error. Ok tapes do go bad and when we looked at the
file it had created up to that point (160 meg uncompressed) it listed more
than 960,000 files. Well 96% is acceptable but when we went into the Xwindow
GUI to do a restore there was still nothing showing.

 After getting bounced around Veritas support for more than a week we were
finally told "sorry but that is the way the system is designed". I was also
told that there may be a way to re-create the index files but that "you
would have to pay Veritas Consulting" to do it. This when we have a support
contract!!! If a fix exists then I feel as a paying customer with a support
contract that I am entitled to it!

The Good, the bad & the ugly:

 The only good thing was that I had not used multiplexing! As such I could
TAR off the files I needed when I discovered this problem. Had I used MPX it
would have been game over. Now TARing files off is very time consuming and
tedious and this is not what we pay Veritas for (and yes they are
expensive).

 The bad thing is that half our classes use MPX and if this happens to them
we are in trouble.

 The ugly thing is that when I asked Veritas tech support how I could
minimize the damage caused by bad media I got a response devoid of thought!
I know if I make more classes and do one 10 or 20 gig volume per class then
I will only lose those files in case of a bad tape. Their advice was to use
their new "feature" Multistreaming which uses MPX!!! They also said that
this wasn't a "support issue" but rather a "configuration issue" and I would
have to pay Veritas Consulting for any other info!

Final Rant:
 If these are limitations of the Netbackup system where in their sales
brochures does it say "if one part of the tape goes bad, you are had"? I
would also suggest not using multiplexing at all if you can help it as it
will destroy your last hope in a situation like this (TARing files off).

Questions:
 
1) How do other tape backup systems handle importing information off of
media that have bad spots? I don't expect to get the info back after the
point in a tape where the media is bad but the stuff before it should be
available for restore.

2) Does anyone know how to recreate the index files under Veritas Netbackup
so it will see the files that are listed in its re-created data file? 

In Netbackup there are 2 files created when you scan a tape or do a backup.
They are put in a directory: 
 /usr/openv/netbackup/db/images/<servername>/xxxxxxxxxx 
The files are <class_name_xxxxxxxxxx> and <class_name_xxxxxxxxxx.f> (the .f
file being the big one). In the <class_name_xxxxxxxxxx> file there is a
NUM_FILES which shows 0 but changing it to 960,000 doesn't work because you
also need to update the files in the directory
 /usr/openv/netbackup/db/images/<servername>/INDEX
That is where I need the help.


Thanks