Re:
> Can anyone please tell me how I can kill a process if it does not want to abort using the abortjob command?
>
> (I must add that this is a batch job with a H U G E sort)

In general, you can't.  I.e., if you've tried :ABORTJOB, then the chance
that any process-kill technique will work is nearly 0.  This includes
the one I introduced in 1990 ... yes, 11 years ago ... in the SHOT tool,
and it includes :ABORTPROC, introduced recently as a new CI command.

Once you've sent a kill request to a process (and, ABORTJOB will do this
for you), sending extra ones won't work.  It's not like yelling at a child
to do something: yelling extra doesn't work.  Hmm...processes are a little
like cats in that respect ... they look at you and say "what *are* you
talking about?"

The "sort" is a key clue.  For reasons that I am unaware of,
SORT seems to choose to "go critical".  I'd say it does this more often
(and for longer) than necessary ... but I don't think a sort should *ever*
"go critical".  When a process is "critical", attempts to kill it are
postponed until the process becomes non-critical.

If you had SHOT, and did:

   :shot
   pin 123              (where 123 is the PIN of the process in SORT)

you'd probably see something like:

   ...
   CRITICAL    :         2  (or some positive value)
   ...
   KILLER      ###
   ...
   PENDING     :             Int: [2]     |

The "KILLER" will be non-0 if SHOT's KILL command was used, or if the
:ABORTPROC command was used, and will probably be 0 if :ABORTJOB was used.

The "PENDING ... Int: [2]" tells me that a level-2 interrupt is pending
against the process ... that's the interrupt level used by a process
kill request.

If/when the process becomes non-critical, the pending kill request will be
handled, and the process will abort.

In short, you have two things to do:

   - wait for the sort to become non-critical

   - while waiting, use the opportunity to logon to the HP ITRC and
     file a bug report against SORT: it shouldn't use setcritical!



Stan Sieler                                           [log in to unmask]
www.allegro.com/sieler/wanted/index.html          www.allegro.com/sieler

* To join/leave the list, search archives, change list settings, *
* etc., please visit http://raven.utc.edu/archives/hp3000-l.html *