Workstation 9 - ESXi 5.1 guest and 10GBe NIC

[COS]

OK, I have a beefy machine running Windows 2012 Server and VMW Workstation 9. The machine has a dual port 1GB NIC and a dual port 10GBe NIC.

All CPU features are enabled in the i7 3820 CPU, so all is good running 64bit stuff.

I have 2 esxi VM's connected to the Automatically Bridged 1GB NIC. The virtual NIC's are "el1000" are in a vCenter Cluster, 2 nodes of course.

Each esxi is connected to an iSCSI target with a 200GB LUN.

Within each of the 2 esxi VM's I have 1 Windows 2008 R2 VM's, one in each esxi VM. Both Windows 2008 R2 VM's are in the iSCSI LUN.

The 2 Windows 2008 R2 VM's are connected to the vswitch of each esxi host, the 1Gb vswitch.

Tthose Windows 2008 R2 VM's fire up fine. Performance is as expected because they are only connected to the iSCSI LUN at the 1GB NIC.

So that configuration works as expected.

Since that worked, I decided, let's use the 10GBe NIC. So I unmounted all iSCSI LUNS and removed the iSCSI target from the iSCSI adpter in both esxi VM's. I'm back to a "Raw build" if you will.

Since I did not have an SFP+ DAC connection to a 10GBe switch, I put an SFP+ Loopback and everything was now pingable on the 10GBe port.

Since my target was presenting the LUN to a 1GB NIC/Subnet, I redirected the target LUN to the subnet of the 10GBe NIC/Subnet.

Everything was pingable on the 10GBe. Awesome right?

Then, I added a new vmnic6 in workstation 9 in "Virtual Network Editor" and "Bridged" vmnic6 to the 10GBe NIC. Next, I added a NIC as a "vmxnet3" to each esxi VM, then configured a vswitch in each esxi VM. No problem there. I see a "10000 FULL" synch/connection. Cool.

Next I configured the iSCSI initiators in each esxi VM to connect to the 200GB iSCSI LUN. The refresh occurs and both esxi VM's see the LUN and it mounts.

At this point, all connections are there and the iSCSI is connected at 10GB/s.

So I import the VM's that reside in the iSCSI LUN, one in each esxi VM. They import fine.

So I power on one of the Windows 2008 R2 VM's that is hosted on the first esxi host and...........PSOD!!!!!

I attached the screenshot.

Can anyone make anything out of this?.........lol

I can repeat this everytime. :smileylaugh:

I laugh because I am probably using WS incorrectly or something.

I did however flash the 10GBe NIC to it's lates BIOS from QLogic/HP.

Thanks

[admin · Accepted solution]

Do not use the vmxnet3 adapter for the ESXi VM.  The ESXi 5.1 driver for vmxnet3 as a "physical" NIC has some issues.  Use e1000 instead.

Note that you can still use vmxnet3 for any nested VMs, without any problems.

[COS]

Welllll

That explains it. What about vmxnet2?

Any idea when there will be a fix?

Any alternative to get to 10GBe? (OK, I think Hyper-V might work but no thanks)

My goal is to hit 10GBe iSCSI backend to test VDI in a "Ghetto Lab". ......lol :smileysilly:

[admin]

COS wrote:

That explains it. What about vmxnet2?

I do not believe there is a vmxnet2 driver for ESXi.

Any idea when there will be a fix?

Sorry.  VMware policy is to not comment on unannounced features, products, timelines, etc.

Any alternative to get to 10GBe? (OK, I think Hyper-V might work but no thanks)

Even if the vmxnet3 driver worked well, I don't think you'd see that kind of bandwidth in a nested environment.  Nested I/O is just too slow, at least for now.

[admin]

COS wrote:

Any idea when there will be a fix?

You might want  to try ESX 5.1 update 1 (Build 1065491) which was just released.

[COS]

I'll bring down the iso and give it the "good 'ol College try".

[COS]

Still get the PSOD..... :smileycry: