Teaming NIC's - no load balancing without switch-side config?

[booradley201110]

Hi folks -- I added a second vNic (associated with a pNic that I uplinked to our Cisco 2960 switch) to a vSwitch, and selected "teaming" and chose the default load-balancing option:  "route based on originating port id".

However I don't see any traffic flowing out (or in) this newly added NIC, even though both are marked as active.

Is some sort of switch-side configuration, like ether-channeling, always mandatory to get load balancing with NIC teaming?

Thx!

[weinstein5]

The load balancing configuration on the ESX is really more of a load distribution - first let me describe the 3 types of load distribution available -

Port Based - This as you found is the default - each virtual NIC or vmkernel port is assigned a virtual port id and based on that id a physical NIC is selected for outbound traffic - so a VM will only use a single physical NIC - there is no configuration on the physical switch to support this

MAC Adress Based - this is a legacy mode carried over from ESX 2 - a physical NIC is chosen for the VMs outbound traffic based on the virtual MAC Address - so a VM will only use a single physical NIC - there is no configuration on the physical switch to support this

IP Hash - Every packet is examined as it being sent by the VM - based on the originating IP address and destination IP address a physical NIC is chosen for the VMs outbound traffic - so a VM's traffic can go out any of the physical NICs requiring the pjhysical switch to be configured for LACP -

None of these methods looks at existing traffic patterns or existing loads in their determination as to which NIC port to use -

So in your example if this was an available NIC that you just added to the switch none of the VMs would have switched to use it since they were already assigned to the first NIC - you would need to power cycle the VMs to cycle through both NICs - or if you had a single VM it would only use a single physica;l port

[booradley201110]

Thanks David, that's an extremely helpful answer, and I understand now....

[rickardnobel]

David Weinstein wrote:

IP Hash - Every packet is examined as it being sent by the VM - based on the originating IP address and destination IP address a physical NIC is chosen for the VMs outbound traffic - so a VM's traffic can go out any of the physical NICs requiring the pjhysical switch to be configured for LACP -

Good explanation on the different NIC teamin options! Just a quick note on the above, since the ESX/ESXi vSwitch does not support LACP the physical switch has to use a static etherchannel / trunk and no dynamic negotiating as LACP does.