Half of a pools VMs compose without a NIC driver

Shoetung · ‎05-25-2011

Hi All,

I have a windows 7 linked clone pool. When the VMs are created half of them generate an error in VMware View (4.6). The error is "no network communication between the VDM agent and the connection server".

Upon further inspection, the machines generating the error do not seem to get the driver for the nic (tried composing with both E1000 and vmxnet3 and the issue occurs with both).

If I log in via the console, and right click the generic "Ethernet Controller" and update the driver Windows will find the driver automatically and it will have network connectivity.

My question is though, why are half of these machines failing to find the ethernet driver for the NIC, yet the other half are fine?

It's rather similar to the issue in http://communities.vmware.com/message/1589696#1589696 however theirs was in View 4.0 and wasn't really resolved.

Can anyone recommend any troubleshooting/solutions?

Cheers

kgsivan · ‎05-25-2011

Whether other half got recomposed successfully, or were pending reompose ?. Please confirm this.

Typically such issue should happen for all if it occures.

What is the ESX version ? and which Guest OS are you using ?

AndreTheGiant · ‎05-25-2011

The virtual NIC is set to use DHCP? Have you release the DHCP before make the master snapshot?

Andre

Andrew | http://about.me/amauro | http://vinfrastructure.it/ | @Andrea_Mauro

npeter · ‎05-25-2011

How many VMs do you have in these pools? Do you have enough esx servers and resources (Memory/cpu) to run all your VMs?

-noble

-nObLe

cwebb00 · ‎05-27-2011

We have experienced this same issue. We had 4 different Windows 7 parent VMs all using the VMXNET3 driver (linked clone pools). The pools had run for at least a month with no problems. Beginning May 23/24 when we took new snapshots of the parents and recomopsed the pools, we found that SOME of the clones were completely missing their virtual NICs inside the Windows 7 guest OS. It was intermittent and after we restarted the parents a few times we saw they began exhibiting the same behavior. Only recent change was that we pushed the May Windows OS Critical and Security Updates. Ticket was opened with
VMware support, but they are stumped at this time. I can only suspect that a recent Windows update may have caused this issue for us. To get the pools working again, we reverted them back to the previous snapshot. Next, I removed the VMXNET3 adaptor and installed an E1000 adaptor. We powered the parents down, took a new snapshot and recomposed off the new one. The errors disappeared. For now, we are going to run with the E1000 adaptor until VMware can assist and figuring out what has broken VMXNET3. Thus far, we have only seen this issue with Windows 7 Guest OS specifically.

Shoetung · ‎05-29-2011

@skg, The other half of the pool recomposed successfully. I would have thought that since all of the VMs were made from the same parent VM that they should either have all worked, or none of them should have worked. Running ESX 4.1 Classic, Windows 7 32bit SP1 (with recent patches).

@Andre, VMNIC set to use DHCP, and IP /release and /flushdns before snapshot.

@npeter, plenty of resources (both mem/cpu), and this pool only has ten in it (5 failing, 5 good).

@cwebb00, Its good to see that I'm not the only one having this issue. Windows Updates may be it. I've done various reinstallations of combinations of agent/VMtools as well as delete/readd NICs, and I think I may have it stable atm, but I'm not convinced that it will remain that way.

kgsivan · ‎05-30-2011

Could you please take the logs and attach ?

Shoetung · ‎05-30-2011

I have attached the View Composer logs and the Agent logs (not sure if this is handy or not) of an affected VM.

I'd attach a log bundle from the View Composer but that is huge.

kgsivan · ‎05-31-2011

"Insufficient disk space on datastore 'R_FCR5_VDI02"

Please ensure the storage has enough free space so that all the VMs can power on and complete the customization.

Also please chek that your active directory does not contain any duplicate machine account name for the VMs which are failing

Shoetung · ‎06-01-2011

I’ll admit I did run out of space a while back, however the two windows 7 pools that are still having these issues were not being provisioned to that LUN, they were to a LUN named L_FCR1-N_W7_01, and another R_FCR1-P_W7_01, both of which currently have at least 566GB free.

LeftHandVSA · ‎06-24-2011

Just to let everyone know, I have run into the same exact problem. My Windows 7 64bit Parent VM is configured with VMXNET3. All the linked clones come up without NIC's. I changed the Parent VM to E1000 and they work like a charm. Oh Well.

vSphere 4.0.0 build 261794