ESXI 4.0 WinXP virtual machines freeze after installing MS security updates

[bbowman]

I have an ESXI 4.0 host server which hosts about 60 VMs running Windows XP sp3.  I have modified the local WSUS policies on each VM which check for and install any available MS security patches every weekend.  If patches are available, they are installed and the VMs automatically reboot afterwards.  My problem is that at least half of the VMs hang while shutting down and are unavailable via RDP.  The only way to revive these VMs is restarting them from the console.  Sometimes the console is just a blank blue screen, other times it says "Windows is shutting down".

It doesn't happen every time with the same VMs, sometimes they reboot just fine, other times they hang while trying to reboot.  Anyone have any suggestions as to why this is happening?  Manually rebooting 30-40 VMs is time consuming, I'd rather they reboot normally like they should.

thanks in advance,

Ben

[vmroyale]

Have you checked the storage to see if all those reboots are overwhelming it with IO?

[bbowman]

I will look into that, thank you for the suggestion.

However on the same host there are about 15 Windows 7 VMs.  These never hang when rebooting after applying MS security updates.  If the host were overloaded by all the reboots, wouldn't the Win7 VMs also hang?

[vmroyale]

15 simultaneous reboots might not be enough IO to cause high latency on your storage, where 60 could be?

It depends on a variety of things though. Are all of these VMs on the same backend storage or how are they spread out?

[bbowman]

All of the VMs (WinXp & Win7) are configured to install available Windows updates every Saturday at 2pm.  I can see how 75 VMs applying updates and attempting to reboot can be a strain on the host's resources but i find it odd that the Win7 VMs never hang on reboot.

All VMs are located on the same datastore which resides on internal storage on a Dell PV R900 server.

I think I'll change the WSUS configurations on 10 VMs to run at a different time than the rest.  If the test group of VMs can reboot on their own then your suggestion about the hardware being overloaded might be correct.