ESXi 4.0 u1 crashing when starting SBS2008

[mudtoe]

Hi folks:

I'm having an intermittent crash problem with ESXi 4.0 u1 when I start SBS2008. It seems like I can start it once, but if I shut it down or reboot, ESXi crashes. After that I can start it again, just once I think (haven't verified this, but the three crashes were on second boots and the times it's worked have all been on the first attempt). The crash occurs right as SBS 2008 is attempting to identify the CPUs (if you use /sos in the SBS 2008 boot sequence you can see the drivers loading, and the crash always occurs after the screen changes from the drivers loaded via bios to when the system drivers start to take over. This is where it lists the windows build version, followed by the number of CPUs and the amount of memory. The crash always occurs after the first line comes out on the console identifying the windows build version, but before the line about the CPUs and memory can come out. Normally these two lines come out really close together in time). The host reboots, so I haven't been able to see if any diagnostic information appears before the reboot. Is there a log kept someplace that might have such information? Also note that this is a Dell 2970 with the two Amdahl 6 core processors in it. The SBS 2008 VM was migrated from a Dell 2950 with Intel CPUs, if that matters. I don't see anything in device manager in the guest that would indicate missing chipset drivers or such (e.g. devices with a "!" next to them). Also, and again I don't know if this matters or not, but I migrated the VM by using the SSH console and the "scp" command to copy the VM's files directly from an ESXi 3.5 instance to the 4.0 instance, and then told ESXi 4.0 that I had moved the VM when it asked the first time I started it up. In all other respects the VM seems to be working fine.

Thoughts and suggestions welcome.

mudtoe

[AndreTheGiant]

Are you sure to have good RAMs on your hosts? Have you tried to run memtest?

The change from Intel to AMD must not cause this problem, if in both case the SBS was a VM.

Andre

[mudtoe]

I think I figured it out. I had a floppy disk drive defined to the VM with a .flp file assigned to it. When I copied the VM to the new server I didn't notice that the device was still defined from when I first installed SBS2008 and needed to put an answer file on a floppy for the install process. In any event the file pointed to by the floppy device didn't exist on the new server, plus it was specified in absolute terms (i.e. /vmfs/volumes/hex string/filename) instead of with the datastore alias name; and somehow that apparently caused ESXi to crash sometimes when the VM was started. Evidently the error checking at run time isn't so good in ESXi, as it must be depending on the editing that occurs when you use the GUI to prevent bad data from getting into the config file. Since I copied the VM directly the editing that would have prevented a non-existent filename to be specified got bypassed. However, I have no good explanation as to why sometimes the VM powered up OK and other times the host crashed; but once I removed the floppy device (I didn't need the answer file anymore), it hasn't crashed since, and I've stopped and started the VM about a dozen times since then just to make sure. I suppose that there must be a "safer" way to migrate VMs from one host to another, versus how I did it, which checks for all that kind of stuff.

mudtoe