beeper transfer & pending process fix

Kate Scholberg (kate@riscgs1.lngs.infn.it)
Sat, 3 Jun 1995 11:02:19 +0200 (DFT)

Hi all,

This mail is to announce 2 things:

1. The beeper has now been transferred to Alec. Erik FedExed it
to Bloomington and it arrived on Friday.

2. I think I have fixed the problem with excess RCD processes
occasionally getting appended to the queue by SENTINEL.

For those interested in the details of #2:

SENTINEL uses the process-checking System Service routine GET$JPIW
to look for processes running the RCDGC_UVAX images. Sometimes this
was failing, apparently at times when VXMACB was running CPU-intensive
jobs -- possibly, this was because GET$JPIW does not see swapped-out
processes. SENTINEL would then launch excess RCDGC_UVAX processes
which would get stuck as pending processes.

To try to fix this, I found another Vax System Service, GET$QUIW, to
return information about whether or not the RCD jobs are sitting on
the queue. This looks for information about jobs running on a queue
rather than information about executing images as GET$JPIW does.

I'm not completely sure that GET$QUIW will always correctly tell
SENTINEL that an RCDGC job is on the queue even if the RCDGC job is
swapped out or VXMACB is having memory problems or whatever (the docs
don't seem to say much about these kinds of things) -- but I think it
will be more reliable than GET$JPIW. The reasons for my belief in
this are:
1. I think info about jobs on the queue is a different kind
of thing from info about running images, and is probably more stable
against weird stuff going on,
2. the command SH ENTRY, which probably
calls something like GET$QUIW knows about all entries on the queue
even when programs calling GET$JPIW don't know about the RCDGC_UVAX
jobs, and
3. GET$QUIW seems to know about holding, pending, suspended,
etc. jobs on a queue, according to the docs and my tests.
Anybody
who's an expert on this kind of thing, please confirm or deny this
belief if you can.

It's hard to test whether this will work for sure, since the problem
is so intermittent, but I think it will work. I'll be keeping an eye
on the queues to make sure things are working as expected.

Kate.