12-15-2011 11:24 AM
I'm in the middle of a project migrating hosts and storage off McData fabrics onto Brocade fabrics. I have a set of seven boot-from-SAN Windows 2008R2 servers that were migrated several days ago to the Brocade SAN.
This morning these servers were all rebooted but three of them would not boot from SAN successfully. The boot LUN was available, but the server blue screens shortly after the OS starts to load. I did get these servers back online by reconnecting them to the McData fabrics. Obviously, this is not an ideal long term solution. The hardware on these servers is identical as is the QLogic driver version, 18.104.22.168. The adapter is a QLE2340. the boot LUN is on our IBM SAN Volume Controller.
12-16-2011 07:06 AM
I'm not sure what else there is to do if I can see the LUNs assigned to the host when looking at the HBA via Ctrl-Q at POST time. Since the server begins to load the OS I know the boot LUN is found. We do see the Win2K8 "Loading..." graphic for a few seconds prior to the BSOD.
I'm thinking out loud here, but this may be a driver issue. The boot BIOS find the boot LUN and begins the process of loading the OS. But when the Win2K8 driver for the HBA is loaded, the server loses the boot LUN or "flips" to the second (data) disk to continue booting. Something is different about these three servers.
Fortunately, the Wintel folks are allowing me to play with one of the servers today, so I'll have more on this later.
12-16-2011 08:42 AM
Many times with SAN boot servers it's an issue of the OS sorting out which path to find the lun on. Try zoning down to a single physical path and rebooting your server. You may have to 're-find' the lun for the OS, this is differnet for the various OS'. You may also need to modify whatever multipathing software you're using, update, clear old path definitions out, etc....
As soon as it consistently boots off of that single path you should be able to add the others.