With our 3590 tape drives rendered "unavailable" to AIX and the 3494 yesterday
we upgraded tape drive microcode, then brought the drives back to service.
However, ADSM for some reason thought one of the drives was "busy", as
revealed through SHOW LIBRARY, though it could readily be used via the 'mtlib'
command, etc. - so a bogus status.

A common way to clear such a condition is to shut down and restart ADSM; but
that would have ruined client jobs that were then running and which we wanted
to complete.  I tried the AIX sequence 'rmdev -l rmt2', 'mkdev -l rmt2' but no
change.  I made the drives unavailable and then available again at the 3494,
to no avail.  I tried power-cycling the drive, but no difference.  SHOW
LIBRARY displayed "polled=1", so ADSM should have seen the true drive state
after all that.  What I finally thought to do was bounce the lmcpd running on
AIX, through which the host communicates with the robot, in case some status
latched in there: kill the process, then run 'cfgmgr' to have AIX bring it
back.  That cleared the problem, and the client sessions took off.  (Just a
'cfgmgr' alone might be enough - I'll try the next time I encounter this.)

A technique worth remembering.     Richard Sims, Boston University OIT