Re: About dispatching process
- From: d10jhm1@xxxxxxxxxx (Jim Mulder)
- Date: 25 Oct 2007 00:08:06 -0700
as far as I know the restart interrupt from the HMC with options
abend-resume-repair are supposed to correspond to the spin(resume)-
retry(repair)-term(abend) options of automatic exspin recovery.
No, there is no connection. Abend/Resume/Repair for the restart
interrupt key was introduced in MVS/370 SP1.3 as part of a 3033
RAS and performance enhancements package (the 3081 had slipped,
and we had to revive the lagging 3033 sales). The Repair option is
a collection of things which never proved to be very useful.
This is what Repair is supposed to do:
Causes limited system diagnosis and repair.
- Call Supervisor Analysis Router (IEAVESAR) to
check and repair critical data areas.
- Look for system non-dispatchable and schedule
an SRB routine to reset the condition.
- Look for WTO queue element depletion.
- Check for system out of work (no batch and
no TSO users), and inform the operator with
message IEA501I via WTO.
- Schedule SRB for IOSVRSTS to analyze I/O.
- Return to interrupted work.
And this is what IOSVRSTS does:
This module is scheduled by the restart FLIH extension module to
run under an SRB. It does the following:
.. Redrives all connected subchannels by invoking the IOSINTRP
macro to simulate an interrupt via a LPSW, which in turn
causes the subchannel to be re-driven.
.. Scans the MIH message queue and notifies the operator,
through the disabled console communications service,
that MIH conditions exist on paging devices.
If IOSVRSTS finds more than one message for
the same condition on the same device, it notifies
the operator only for the first occurrence of the
message. If the disabled console communications
routine fails, IOSVRSTS stores the device number in
the PSA (PSAWTCOD) via the LOADWAIT macros used by the
associated Loadwait/Restart service which places the system in
a restartable wait state with a hexadecimal wait state code of
116.
.. Terminates any address space in which I/O purge is waiting and
the purge quiesce counter is not equal to zero. IOSVRSTS
terminates the address space by issuing CALLRTM TYPE=MEMTERM
with a hexadecimal abend code of 0AE.
I was the function tester for that whole RAS package in SP1.3.
The item to memterm address spaces which are waiting in
I/O Purge was rather controversial among the development team.
It seemed to me like a bad idea, but I didn't have much influence
on the designers 27 years ago. It still seems like a bad idea,
but since I doubt that this Repair function ever got used much
outside of my testing, I haven't bothered to get it deleted.
(I do have a bit more influence now).
Automatic Excessive Spin Recovery did not come along
until MVS/ESA SP3.1.0.
Jim Mulder z/OS System Test IBM Corp. Poughkeepsie, NY
----------------------------------------------------------------------
For IBM-MAIN subscribe / signoff / archive access instructions,
send email to listserv@xxxxxxxxxxx with the message: GET IBM-MAIN INFO
Search the archives at http://bama.ua.edu/archives/ibm-main.html
.
- Follow-Ups:
- Re: About dispatching process
- From: Rob Scott
- Re: About dispatching process
- References:
- Re: About dispatching process
- From: Barbara Nitz
- Re: About dispatching process
- Prev by Date: Re: Outsmarting WLM
- Next by Date: Re: About dispatching process
- Previous by thread: Re: About dispatching process
- Next by thread: Re: About dispatching process
- Index(es):
Relevant Pages
|
