> Tape Library: IBM 3584 partitioned into production and test 
> environment
> 
> Operating System:  Solaris 9, Windows 2003, HP-UX 11i
> 
> Netbackup: 5.1 mp3
> 
> Just upgraded from STK L700E LTO-1 library to a IBM 3584 LTO-2 library
> 
> All of a sudden tape compression is not working on any flavor 
> media server,=
>  but only in production. Our test environment is compressing 
> data at a hard=
> ware level just fine.  Same OS, same NBU release, same Tape 
> library, same t=
> ape drive drivers, same HBA cards, same firmware, same 
> st.conf file on the =
> Sun systems.

Have you consulted the NetBackup Hardware Compatibility List?  Both the
5.1 and 6.0 lists show "Library partitioning is supported" for
TotalStorage 3582 and 3583 models, but the 3584 models say "All models
supported. --- ALMS is supported with the exception of drive sharing."
No mention of library partitioning.

Un-partitioning the library for a test should show if this is germane.

> color:blue;font-weight:bold'>Although the spec is the
> same, do any of the netbackup logs show why compression
> may not be running?</span></font></i>

AFAIK, NetBackup knows/does absolutely nothing about compression--it
uses the drive/driver that you configured.  If you configured
/dev/rmt/0bn, you gave it an uncompressed drive[r entry point], if
/dev/rmt/0cbn, a compressed one (assuming the name matches the maj/min
numbers, of course).  You know that already.  The point is, NetBackup
sends the driver a write command, a buffer address and a count and gets
back a good/bad status from the driver at the end of the operation.
What the hardware actually put on tape is up to the hardware and the
mode to which its driver set it.

In case you carried over your old media to one of the new environments,
you know that LTO-1 media in an LTO-2 drive is written to at LTO-1
density.

Update from digest #5102:
> I answered the compression issue in a previous post.

Apologies.  You did, but words are sprinkled throughout pages of
impenetrable HTML crap.  I would still recommend a) setting up a simple,
repeatable test as suggested and b) proving that the partitioning of the
library is not a factor.  Now, I'd add c) involve or eliminate NetBackup
by using dd, rmt, fdump or other utilities to see if they exhibit the
same behaviors you see on your test/prod sides.

You have completely ripped out the old lib/drive configuration (dev,
fibre, st, sg, NBU) and rebuilt it with the new hardware, correct?