Subject: | |
From: | |
Reply To: | |
Date: | Wed, 8 Jul 1998 12:13:25 -0700 |
Content-Type: | text/plain |
Parts/Attachments: |
|
|
On Jul 8, 9:27am, Gavin Scott wrote:
> If the implementation of multiple job queues looks the way we discussed
> on the list last year, then it is effectively nothing more than an
> automatic :LIMIT management system for controlling which jobs are
> allowed to start. Once a job starts the MJQ functionality won't have
> any further effect on it. So "hidden" would be an attribute of an
> executing job, where as MJQ concerns itself with waiting jobs.
Gavin is essentially correct and you can read more on MJQs on the jazz
webserver. As Mike B. points out you can move an active job from one queue
to another but this does not impact the moved job in any way. It may have
been nice to have a workgroup (Workload Mgr) job queue property so that the
job queue could impact the exection of its jobs.
I recall a discussion about hidden jobs at SIGMPE (?) but
I don't remember if it made the SIG list. One of the motivations for
hidden jobs was to not impact the job limit for "normal" visible production
jobs. I think MJQ may help here. Another motivation was to really hide the
jobs so that someone wouldn't accidently abort it (or something like that).
MJQ doesn't help for this need.
> If :SHOWJOB is enhanced to select jobs based on which queue they launched
> out of, then you could change your :SHOWJOB commands and UDCs to exclude
> things that came out of the "hidden" queue.
SHOWJOB does not filter based on job queue name, although this would be nice
(and so would general selection equations). The new ;JOBQ parm for SHOWJOB
causes the output to reveal the job queue name.
Jeff Vance, CSY
--
|
|
|