Hi,

This is probabkly more of a Solaris issue than NBU, but I was hoping
one of you guys might have seen it before...

We have an E450 running Solaris 2.6 using Veritas Netbackup to
back up to a Storedge L9 DLT library. The drive and loader in
the library are daisy-chained on the same SCSI bus, the loader
has ID 0, the drive ID 1 and there are no other devices connected
to that SCSI bus.

The robot in the L9 failed, and the unit was hot-swapped out by Sun
without rebooting the box (it's a 24x7 box, rebooting is not
really an option here).

The new unit appears to function OK:-
* Inventories itself fine and passes its "wellness" test
* Veritas can identify the DLT drive and seems to
  communicate with it OK
* Veritas can identify the loader but *not* communicate
  with it

We don't think this is a Veritas issue, as even running their
low level tldtest command fails:-

# /usr/openv/volmgr/bin/tldtest -r /dev/sg/c2t0l0
Opening /dev/sg/c2t0l0
mode_sense failed, CHECK CONDITION
sense key = 0x2, asc = 0x4, ascq = 0x1, LOGICAL UNIT IN PROCESS OF BECOMING 
READY

Things that have been tried/ruled out:-
* Dodgy SCSI chain - Both the drive and loader show up fine in sgscan:-
/dev/sg/c2t0l0: Changer: "HP      C7145-8000"
/dev/sg/c2t1l0: Tape (/dev/rmt/0): "QUANTUM DLT8000"

* Power cycling the library (it comes up, invetopries itself and carries
on giving the same responses)

* Re-running drvconfig, devlinks, sg.build etc


So far Sun's only suggestion has been to reboot the box which, as I 
mentioned, would not be an option for some weeks.

Any other ideas? A Google search suggests this has happened to other
people, but I've not managed to find an explanation...

Pete.