Entertaining myself here at lunch, I present, in no particular order...

10. Can't set exclude lists at the master, only at the client.  Very silly.
I would like to exclude, say, /tmp or WIN386.SWP on all clients in a class
but I can't...instead, I have to make sure the exclude file is correct on
each client.  What a nightmare for large environments.  Some admin adds a
client (that I may not have access to check his setup) and doesn't set his
exclude list properly.  Now I'm chewing tapes for nothing, or worse, not
backing up something he needs.  What was the thinking here?  Particularly
since the raw Class database shows an Exclude field ;)

9. "Use any available media server" doesn't round-robin.  Strange but true -
it works the list of media servers *alphabetically*.  If you have multiple
media servers, then NetBackup will attempt to use all of the drives on #1,
then go on to #2, etc.  If one of the drives in #1 becomes free, then it'll
use it before going back to #2.  So if you have multiple media servers and
want to distribute the load, it's difficult to do so because your full tapes
will naturally accumulate in the first (alphabetically speaking) storage
unit.  

8. Can't restart just one stream.  If a single stream fails, you have to
rerun the whole backup job.  Blech.  Why can't I restart just that one
stream?  

7. The X-windows backup/restore GUI doesn't have a "preview media required"
feature for a restore.  The Java backup/restore GUI doesn't show you a list
of the backup images.  I wish there was one that did both ;)

6. bpinject sucks.  I think bpinject is only packaged with bpvault.  The
design is really amazingly bad.  The idea is that you can type "bpinject 2"
and it'll dump the cap from robot 2 into the robot.  However, if that robot
is on another machine, then you need to have root-level .rhosts access to
it.  Oh sure, let me rewrite our site security policy so I can insert tapes
- har!  This is a particularly bad design because you can use Netbackup's
vmchange to insert the tapes through NetBackup.  The only reason you need
.rhost access is to query the robot - I'm opening my whole system up just so
you can query the robot!?!?

A better solution is simple:

        - write a service on the media server that runs out of inetd that
reports the status of the robot.  Takes
        about 100 lines of perl.
        - when the bpinject script needs to know what the status of the
robot is, it can connect to this service 
        and query the status of the robot, then use vmchange (vmadd, etc.)
to insert tapes.

5. jnbSA sends the root password in the clear.  Come on people, this is 2002
- how hard would it be to encrypt this?

4. Frequency for hours, days, and weeks...but not months!  How can I do once
a month?  I can do every 4 weeks, but that doesn't really work because
there's an extra week every quarter. 

3. Stupid bug - put in a path like "/somewhere/someplace/blah   ".  Note the
blanks.  Now run a coverage report.  Note that this is reported as
"UNCOVERED".  Sheesh.  Also, if you put "C:" in a file list, you'll get only
the "Veritas" folder on C:.  If you put "C:\", though, you get everything.  

2.  Archive doesn't make two copies, only one.  Considering that you're
blowing the data away from disk, I'd feel better if it was written to two
tapes, but NetBackup only does one.  I wish this was configurable.

1. It's still better than any other product on the market.