High percentage of dropped packets in ESXi 4

Have you checked for a speed or duplex mismatch between the 1000v port and the vm in question?

Yes, they are all at 1gb... its happening to almost all VMs in the envirnoment.

Take a look at http://www.ivobeerens.nl/?tag=vmware

The article concerns itself with IPv6 being enabled on Windows 2008 guests, and both disabling IPv6 (not a good idea in 2008) or changing the NIC type to vmxnet3 fix the problem.

If this doesn't work then more information is in order to troubleshoot the problem.

1. What guest OS types are involved (include 32/64 bit information)
2. What are the virtual NICs? E1000, flexible, vmxnet2, vmxnet3, etc.
3. Is vmware tools installed and current on all of the troublesome guests

Good Luck

The ipv6 is enabled. However, I just disabled it and restarted the server but it still has the similar packet loss... %drptx is 0 as before...
I am running windows 2008 R2 (mostly)
e1000
and yes the tools are insatalled on all the VMs and are upto date.

I would probably change the NIC to vmxnet3. To do that you must delete the old NIC, then add a new NIC specifying vmxnet3.

It is best to leave IPv6 enabled on 2008 and 2008 R2 because some services don't behave properly without IPv6 to bind to (or so I have been led to believe) and it really shouldn't hurt anything.

The vmxnet3 NIC will often give considerably better network performance than the e1000

Once I change the nic to vmxnet3, i see 0 drprx... thats good .. however, what are the drawbacks of using this vs e1000?

I don't know of any drawbacks - vmxnet3 supports everything E1000 does and is better "paravirtualized" so there is less overhead and better performance. It is generally recommended to switch to that.

interesting! so are there any limitations to this? DRS fails, vMotion etc etc? Any funtionality that it takes away. I guess this cant be used on a FT enabled VM.

Absolutely - it can be used for all of the above