Conference wonder::turbolaser

1192.0. "ASE fails on DEFPA after crash" by BIRMVX::RHODES () Thu May 01 1997 10:57

 
Hi,
	We have a pair of 300 MHz AlphaServer 8200's running unix V3.2D and
configured as an ASE cluster and the system has been SOLID for over 1 year.
	 
The customer has been sold some 440 MHz cpu's, so last week we 
scheduled some downtime (in the early hours of the morning, it's a production
system! ) to upgrade the firmware and then ultimately the operating system to
V3.2G to support the 440 MHz cpu's.

	Initialy we failed all the services over to one system and upgraded the
firmware on the other to V4.0-4, all seemed to be okay until we power cycled
that system. At that point we had messages about the DEFPA  link being
unavailable, well sure enough we booted the system and and FDDI link did not 
work. The link consists of a DEFPA-DA module in each system ,PCI slot 9, V2.46
firmware with just fibre cables between the systems. At that point we decided
to upgrade the other system to V4.0-4 in case it was a console firmware
mismatch issue.

	Well with both systems at console V4.0-4 and having been power cycled
together whilst we did get a message on one system about the FDDI link being
unavailable the system did boot and the FDDI link worked, we were able to ping
the other system!.

	With his system back customer would not let us investigate further and
I agreed to research this offline. Well note 1096 shows other people have seen
the problem, only our system is slightly different in that if both systems are
power cycled the FDDI link works...OR SO I THOUGHT!!!!!!!.

	Several days later one of the systems paniced  (a known Lat problem )
but on reboot the FDDI link refused to work and as it is used for ASE
communication the customer could not ping other system and therefore fail
services back. The customer did not document anything at this point, he just
shutdown remaining system, power cycled both and rebooted, both booted cleanly
and FDDI working. A planned call then logged for early hours to downgrade
firmware to V3.0-9 with no time to investigate problem further.

	Current situation is that systems both on V3.0-9, everything working
okay, customer wants V4.0-4 so he can more forward and use his 440 MHz cpu's. 
BUT we cannot have system to further define the trouble statement (ie were leds
on, off or flashing, what happened if you did this? ). As I have said note 1096
has outlined part of the problem (and no solution ) this goes further and I am
putting this note in prior to raising an IPMT.

	Thanks and regards Mike Rhodes