----- Original Message -----
On Fri, 30 Mar 2012, Ayal Baron wrote:
Have you encountered any issues with it? although the bond supports 9000, sending a 1500 packet from the guest should work just fine.
Right, the issue I am having is larger then 1500 packet being sent to the guest.
The only way I can think of to control this per guest would be to configure the tap device of the vNIC, however, all devices connected to a bridge have to have the same mtu so changing one tap device would require changing them all which would beat the purpose. So I'm not sure what you're actually asking for here.
Today what I do is this. Fire up the VM, find out what host its on, figure out the vnet interface and then ifconfig it with smaller MTU if it is say vnet1 I would:
ifconfig vnet1 mtu 1500
This seams to be working, other hosts that have vnet interfaces at 9000 bytes still work and nothing larger then 1500 bytes is sent to the vnet I lower to 1500. I now just need a way to keep from doing this every time manually.
So if I have 3 VMs that require 1500 and I have 3 hosts and they just happen to be scheduled one per host, my other 100 VMs that do support 9000 would be degraded. Instead, separate these 3 VMs to a different vnet and everything just works. When you change the vnet you're degrading all the VMs running on this host using this network.
-Nathan