Joe -

One more approach is to employ a PRESchedulecmd to kick off a
background process which will run alongside the dsmc and terminate it
when its clock window runs out.  In Unix, dsmc will terminate when it
receives one of: SIGTERM, SIGHUP, SIGINT, SIGPWR, SIGQUIT, or
SIGUSR2. (It will also terminate in response to a SIGALRM, but that's
too "meddling".)

As you've seen from prior postings, the product lacks the obvious
capability to perform a Disable Session, to have the effect of
gracefully signalling one or all client sessions to end at the
conclusion of their current transaction.  Existing commands, such
DISAble SESSions and LOCK Node, are effective only to prevent new
sessions.  (See IBM Technote 1169370.)  The closest you could come to
this by "rigging" is to have time-windowed clients back up to a
designated storage pool which was the only or last in a storage pool
hierarchy, to render the stgpool Readonly at the end of the time
window, thus causing the next write attempt to encounter an out-of-
space condition and the session to terminate at a relatively clean
boundary.

   Richard Sims