[HECnet] UPS driven OPC$CRASH?

[Steve Davidson]

APC offered software to detect power failure using their UPSes for VMS but it was for the units that predate machines with USB support. Connection was via serial port. 

-Steve Davidson

SF:iP1

> On May 31, 2020, at 18:46, Johnny Billquist <bqt at softjar.se> wrote:
> 
> On 2020-05-31 21:02, G. wrote:
>>> On Sun, 31 May 2020 14:30:15 -0400, Thomas DeBellis wrote:
>>> Can anybody elaborate more on what the VMS power failure options are?
>>> OPC$CRASH sure looks like a crash, but what happens before the belly up?
>> You may want to take a look at the VAX/VMS Internals and Data Structures
>> manuals on Bitsavers where you could find editions for V3.3, V4.4, and V5.2.
>> There is a whole part (VII or VIII depending on the edition) completely
>> devoted to the bootstrap process, initialization, shutdown, power failure
>> and recovery, and so on :)
>> See http://bitsavers.org/pdf/dec/vax/vms/training/
>> Quote from the V5.2 book: «Powerfail recovery support enables a suitably
>> equipped VMS system to survive power fluctuations and power outages of short
>> duration with no loss of operation. The support is provided by hardware
>> features (battery backup) and VMS software routines. VMS support includes a
>> powerfail service routine that saves the volatile state of the machine when
>> the power fails, a restart routine that restores that state when the power
>> is restored, CPU-specific initialization code, and device-specific code
>> within many VMS device drivers. The VMS software also provides process
>> notification by means of power recovery asynchronous system traps (ASTs)».
> 
> Very similar story as with RSX. However, I wonder if that was kept maintained in VMS.
> It requires that all memory are either core (which don't exist for VAXen), or battery backed. Battery backed memory certainly did exist, but I suspect that eventually went out of fashion.
> 
> Same on RSX. Either core, or battery backed main memory. The CPU volatile state is saved at power loss, and execution resumes when power is restored. Device drivers and programs can get notified about a power loss and resumption.
> 
> Dumping all of memory to permanent storage was never an option, as the power loss means you only had something in the order of 20ms of execution time left.
> 
> UPSes work on a very different principle. They are meant to be able to keep the system running for a while when power is lost. And then you can do a graceful shutdown if needed.
> 
> Not sure if VMS have any hooks in place to force a graceful shutdown. RSX do not, that I can remember.
> 
>  Johnny
> 
> -- 
> Johnny Billquist                  || "I'm on a bus
>                                  ||  on a psychedelic trip
> email: bqt at softjar.se             ||  Reading murder books
> pdp is alive!                     ||  tryin' to stay hip" - B. Idol