Matthias,

The SHOW LIBR command is a good place to start to diagnose this problem.
Here's what the flags you
mentioned mean:

busy=  0 means the drive is not mounted or even acquired by adsm.
online=  0 means the drive is "offline".  In Version 2, this will only be if
the polled= flag is also 1.
             In V3, you can update a drive to be offline, in which case the
polled flag will be 0.
polled=1 This means that ADSM could not use the drive for one of three reasons:
     The drive is loaded with a Non-ADSM volume (eg a cleaner cartridge, or a
volume from the other ADSM server)
     The drive is unavailable to the library manager (usually set this way by
load/unload failures)
     The drive cannot be opened  (some other application has it open, or
there's some connection problem, etc)
The polled=1 means the server is polling the drive every 30 seconds to see when
the above three conditions all clear.
(It also means that the online flag should be 0.) When the conditions clear, it
turns online back to 1 and the drive
should now be available to be acquired.

I would leave your mountlimit set to 6 on both servers for now, and try to
monitor this situation.  When you see a
delayed mount, look at the show library again and see if you can account for
the delay.  If a drive is polled, then it's not
available.  If they are all polled or busy, that is why you don't get the mount.

Keep in mind there ARE other reasons for a delay, most having to do with
waiting for a particular volume.  Also, if one
server's mounts are IDLE, then they are going to be offline/polled on the other
server!  So set your mountretention to a low
value.

One problematic area is that ADSM transactions often call for a series of tape
mounts, not just one.  If
one server dismounts a drive in order to mount a second volume in the series,
the other server is free to grab and use
that drive.  This may force the first server to have to wait for it to finish
before the second volume can be mounted.
For this reason, you need to have a longer mountwait value in your devclass,
because if it cannot get access to
a drive after this value expires, the mount and transaction will fail.

You may find that setting your mountlimits to 4 or 5 is better, because then
you are not sharing ALL the drives and
this will lessen contention.   Setting them both to 3 is not much better than
not sharing at all except that all six drives
are available as a sort of drive pool.  You will have to find the best setting
for your workload.

Craig Bell