I've seen some instability in one of my VMs.
Host environment is ESX4_U2 on a whitebox server an ASUS motherboard and Intel 82573L onboard NICs. The VM is running Windows Server 2008 x64 and is reasonably up to date. VMWare tools is in sync with ESX4.
The box pushes a lot of traffic as its my backup server, and also runs mail archiving and network monitoring software.
Windows has crashed a couple of times since I've P2Ved this system (the guest used to run natively on the hardware until I installed ESX4 recently). I setup the debugging tools and ran a detailed debug (!analyze -v) and got this:
FOLLOWUP_IP:
vmxnet3n61x64a182
fffffa60`00fc5182 413ac5 cmp al,r13b
SYMBOL_NAME: vmxnet3n61x64a182
FOLLOWUP_NAME: MachineOwner
MODULE_NAME: vmxnet3n61x64
IMAGE_NAME: vmxnet3n61x64.sys
BUCKET_ID: WRONG_SYMBOLS
Followup: MachineOwner
-
0: kd> lmvm vmxnet3n61x64
start end module name
fffffa60`00fbb000 fffffa60`00fd0000 vmxnet3n61x64 (no symbols)
Loaded symbol image file: vmxnet3n61x64.sys
Image path: \SystemRoot\system32\DRIVERS\vmxnet3n61x64.sys
Image name: vmxnet3n61x64.sys
Timestamp: Fri Feb 12 09:39:01 2010 (4B756805)
CheckSum: 0001827D
ImageSize: 00015000
Translations: 0000.04b0 0000.04e4 0409.04b0 0409.04e4
One thing that I'm doing that's probably not ideal is that I'm not separating the VM traffic from the service console traffic (i.e. only one of the four NICs is plugged in), but that shouldn't cause these kind of problems.
I'm using the vmxnet3 vNIC because the E1000s were causing performance problems with my network monitoring software. Should I try the vmxnet2 vNIC instead? Anything I can do to prevent these crashes?