[ofw] RE: installation/connectivity problems on hpc server

[Tzachi Dar]

As for the ipoib connecting to the broadcast group issue. Can another
instances of ipoib connect to that broadcast group ? (for example
machines that have the ndis 5.1 version installed?).
 
As for the dump file. We are not able to locate the correct symbols.
What version have you been using? We get the following error: 
 
http://www.openfabrics.org/downloads/WinOF/v2.1/Server_2008-Vista-HPC/Sy
mStor//mlx4_bus.sys/4A16ADB132000/mlx4_bus.sys not found
So we can not get the true symbols. Can you please run !analyze -v and
send us the results.
 
 
As for the system not booting. Can you load it using WinPE and copy the
mlx4_hca.sys file to it. I hope that after that you will be able to
load.
 
Thanks
Tzachi


________________________________

	From: Anatoly Greenblatt [mailto:anatolyg at voltaire.com] 
	Sent: Thursday, November 12, 2009 1:44 PM
	To: Tzachi Dar; Smith, Stan; ofw at lists.openfabrics.org
	Subject: RE: [ofw] RE: installation/connectivity problems on hpc
server
	
	

	Hi,

	 

	We had 2 (and now even more) problems on these systems.

	 

	1st is the problem of installation but we have a workaround.

	2nd is the ipoib problem system log shows: "Mellanox IPoIB
Adapter #3: Subnet Administrator failed query for broadcast group
information."

	3rd two systems had bsod (minidump is attached). After bsod, one
system stops booting and the text screen shows that "system failed to
boot because critical system driver is missing: mlx4_hca.sys

	 

	Regardsm

	Anatoly.

	 

	
________________________________


	From: Tzachi Dar [mailto:tzachid at mellanox.co.il] 
	Sent: Thursday, November 12, 2009 12:19 PM
	To: Smith, Stan; Anatoly Greenblatt; ofw at lists.openfabrics.org
	Subject: RE: [ofw] RE: installation/connectivity problems on hpc
server

	 

	IMHO, if vstat shows the links as up, this is not an
installation problem but rather an ipoib problem.

	 

	Can you please run ipoib with trace and send us the logs (also,
do you have anything in the event viewer)?

	 

	Thanks

	Tzachi

		 

		
________________________________


		From: ofw-bounces at lists.openfabrics.org
[mailto:ofw-bounces at lists.openfabrics.org] On Behalf Of Smith, Stan
		Sent: Thursday, November 12, 2009 12:35 AM
		To: Anatoly Greenblatt; ofw at lists.openfabrics.org
		Subject: [ofw] RE: installation/connectivity problems on
hpc server

		Hello,

		  Please see inline comments.

		 

		
________________________________


		From: ofw-bounces at lists.openfabrics.org
[mailto:ofw-bounces at lists.openfabrics.org] On Behalf Of Anatoly
Greenblatt
		Sent: Wednesday, November 11, 2009 1:05 PM
		To: ofw at lists.openfabrics.org
		Subject: [ofw] installation/connectivity problems on hpc
server

		Hi,

		 

		Should we have any problems installing Winof 2.1 on
server 2008 hpc edition sp2? Or am I missing something. 

		 

		Have not attempted a WinOF install on Svr 2008 HPC sp2,
although I would not anticipate any problems.

		Try installing via 'start/wait msiexec /I
WinOF_2-1_wlh_x64.msi /Lv msi.log'. Grep log file for error.

		 

		The installation ends prematurely claiming that previous
installation was detected. 

		Is this a MSFT installer error message or a WinOF
installer error message?

		 

		I've exctracted the drivers from winof hpc x64 msi and
installed manually. 

		 

		Since you have already installed HCA drivers by hand,
you want to install WinOF with NO devices installed 

		  'start/wait msiexec /I WinOF_2-1_wlh_x64.msi NODEV=1'
# just take the default install, no devices will be installed. 

		 

		The bus/hca/ipoib drivers were installed successfully,
however the ipoib network adapter shows status "disconnected" 

		State of cable disconnected indicates the SM has not
seen/configured --> Active port state. 

		 

		Opsnsm is running on linux node and shows all
ports/nodes as connected. Vstat on the nodes shows that ports that are
physically connected are up.   UP == Active port state? 

		 

		It is c-class hp blade with connectx gen2. firmware
2.6.1.

		 

		Any ideas how to fix ipoib connection?

		 

		Thanks,

		Anatoly.

		 

		 

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.openfabrics.org/pipermail/ofw/attachments/20091112/d039bc03/attachment.html>