On Wed, May 21, 2014 at 5:09 PM, franz9000 <franz9000@xxxxxxxxx> wrote:
I'll be the one to disagree.
It's my choice to put the jobq on hold, for whatever reason.
I don't think that was the question. The *job* was put on hold, not
the queue. In the case of a single held job, the prevailing view is
that not submitting any further instances of the same job makes the
most sense. I agree with this. If the *queue* is on hold (and the
job itself is released), then I probably would agree with you that new
instances should be submitted freely.
Why should AJS make it's own decision (by design) not to submit,
when the scheduler is running and the time arrives? It has no
application knowledge to know if that is a good idea or not.
Of course no scheduler will have application knowledge. But in the
case of one *job* being held, the odds are that not submitting further
instances of that job (even if it turns out that it was safe to submit
them) is less costly than submitting further instances when in fact it
is not safe to submit them.