I think the situation with the 3880's (or was that the later 3830's) was that if DMKRIO did not have
all the control unit devices it would loose interrupts.   I forget the condition, but the control unit
would return an interrupt on the highest available address it was configured for, and if DMKRIO
didn't have those addresses, VM would basically ignore the interrupt.  

Ah, the "good old days"...................... :-)




"Edward M. Martin" <[EMAIL PROTECTED]>
Sent by: The IBM z/VM Operating System <[email protected]>

04/18/2006 11:12 AM

Please respond to
The IBM z/VM Operating System <[email protected]>

To
[email protected]
cc
Subject
Long IPL process was RE: Wishlist item for z/VM 5.3





Hello Everyone;

 

      “Back in the day”,  VM would not run very well if the DMKRIO was not generated to

match what the CE had set in the phyical hardware configuration.  If the DASD controller (3880 iirc) was generate for 64 devices, then the DMKRIO had to be generated to have 64 devices.

 

      Seems that it was the matter of how the VM handled the I/O interrupts and the like and where the

Response was expected.

 

      Could your long IPL be something along the same lines?  The IOCP and the actual hardware devices

do not match?

 

Ed Martin
Aultman Health Foundation
330-588-4723
[EMAIL PROTECTED]
ext. 40441


From: The IBM z/VM Operating System [mailto:[EMAIL PROTECTED] On Behalf Of Colin Allinson
Sent:
Tuesday, April 18, 2006 8:10 AM
To:
[email protected]
Subject:
Wishlist item for z/VM 5.3

 


We recently had a problem with extremely long IPL time. While we have solved this for now with a tidy of the IOCP, we don't know exactly what caused the problem.


It would have been a great help in diagnosing the problem if the IPL process put out messages on the IPL console to say what it was doing and how far it had got (similar to the shutdown process now does).


When I am talking about long IPL time I mean veeerrry Loonngg (i.e. 90 minutes+ in some cases.) While we have a very complex configuration with a very large number of devices this clearly indicated that some sort of MIH was cutting in somewhere - but we had no indication where. We are now back to a respectable 15 mins.


Colin Allinson  

Reply via email to