ESXi 4 VM BSOD if it is the second VM on the LUN

[Philip_B]

Greetings-

Just deployed StarWind 5.6.1659 (console) on a test RAID 10 array and have (2) 256 GB LUNS.

Connected ESXi with a pair of Intel GB NICS and am using round robin MPIO and switching after each IOP. The first VM after migrating with vCenter fires right up and the performance is very solid. Looking at the Starwind box I can see the I/O spread across 4 Nics on the storage side.

But when I start a second VM on either the LUN that is already holding a running a VM or another LUN on the storage box, the new VM bluescreens with a :

STOP: C000021 unknown hard error and references NTDLL.DLL.

This is very consistent as I have tried it with several VMs - the first to start is fine and any after it will all BlueScreen.

But if I shut the first VM down and boot the second VM - it comes up fast and stable. I can only run one VM at a time.

Starwind server has:
4 GB of ram
(6) Intel 1000MT interfaces
(1) Xeon 2.8 Ghz CPU
3Ware 9550-8 raid controller
4 WD RE3 SATA2 hard drives RAID10 (data)
2 WD RE3 SATA2 250 drives RAID 1 (boot)

When I carved the LUNS I set them up for cluster, but right now only a single ESXi 4 box is connected. As a test I connected two physical boxes to two additional LUNS and am able to use them very reliably.

Not sure where to look at this point for what is causing the issue when the second VM starts up.

Any thoughts would be greatly appreciated.

Thanks.

[anton (staff)]

Frustrating... Do you happen to have 1) all latest ESX updated rolled on and 2) can you grab StarWind logs, zip and send us back to support@starwindsoftware.com so we could take a look? As an initial troubleshooting round... Thanks!

[Philip_B]

Thank you Anton-

Logs were submitted via email.

ESX is fully patched and so is the Starwind box (S2008-32 bit). I left the one VM running with a burn-in utility all night - had the CPU, memory and disk maxed and it performed beautifully... so I suspect that I fat-fingered something with authentication on the storage side.

Odd thing is that I get the same response if I put the second VM on the same LUN or a different LUN... very strange.

[anton (staff)]

Could be! Let's check the logs initially... Thank you!

Thank you Anton-

Logs were submitted via email.

ESX is fully patched and so is the Starwind box (S2008-32 bit). I left the one VM running with a burn-in utility all night - had the CPU, memory and disk maxed and it performed beautifully... so I suspect that I fat-fingered something with authentication on the storage side.

Odd thing is that I get the same response if I put the second VM on the same LUN or a different LUN... very strange.