At Piedmont Airlines, we had problems with V/FORCE-XA on XA/SP, and we did not have an Amdahl cpu. It seemed that after using V/FORCE, the system would always crash at some undetermined later time. We found it easier to leave the user in LOGOFF/FORCE PENDING and schedule the outage at our convenience. We didn't like the consequences if we interrupted our executives' ability to use PROFS.
-----Original Message----- From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Phil Smith III Sent: Friday, January 05, 2007 2:45 AM To: [email protected] Subject: Re: Logoff force pending Shimon Lebowitz <[EMAIL PROTECTED]> wrote: >I seem to remember that the old V/Force included somehow disabling all >devices attached to the hung user, so that even if it DID 'wake up', it >no longer would/could do any new I/O. >I think it might have also detached devices not 'in use', but I am less >sure of this. >Anyone remember more details? Uh, yeah, I do...since I wrote V/FORCE-XA (in a week while camped at a customer site during the VM/XA SP Release 1 ESP). Those were the days... V/FORCE-XA did several things: 1) Ran the MDISKs and made them all R/O 2) Stacked CPEBKs to detach all devices, starting with the least likely to cause problems (MDISKs) and proceeding up through the more likely (tape drives, consoles) 3) Stacked CPEBKs to do other cleanup (CSE, etc.) 4) Set a timer for 15 seconds and waited for the CPEBKs to be all processed, checking every second to see if they were gone 5) If the CPEBKs all cleared, set the "force the user off" bit and waited again 6) If the CPEBKs didn't clear, or the userid didn't go away after setting the "force" bit, it renamed the userid (including updating the userid hash table) and force disconnected it from any real tube Mike Walter <[EMAIL PROTECTED]> >I distinctly remember back in 1986 bringing in V/Safe (intercepted >common CP ABENDS, doing fix-ups where possible and letting the system >continue without an IPL), V/Snap (before IBM had a CP SNAPDUMP >command), and V/Force. >When we needed it, V/Force went into recursive ABENDs, taking down the >system (and almost my next salary increase!). John (don't remember his >last name) from VMSG flew in several times for weekend S/A time to >recreate (successful) and resolve (unsuccessful) the problem. Pity. >But now we seldom see the error anyway (hence, my "HUNGUSER HELPME" >since I won't remember all the techniques to diagnose it from memory). Yeah, I remember that. That was John Holcomb, and HPO; the XA (and ESA) versions were MUCH more stable, since the XA operating system code was much better than the HPO code (and the V/SAFE-XA code was also better). You had some weird Amdahl box whose timing was very different from IBM boxes; we were sure that was the cause, but never got past that point in diagnosing the issue. ISTR that this was the only time we ever had to give a refund to a customer. Wow, that all seems like a million years ago...(well, it was twenty or so!) ...phsiii
