Comments on: ESX Server, NIC Teaming, and VLAN Trunking

By: slowe

slowe — Fri, 30 Sep 2011 17:19:23 +0000

Sandip, on VMware ESX/ESXi, NIC teams would not be assigned an IP address. IP addresses would be assigned to Service Console/management ports, VMkernel ports, or specific VMs, but not to the NIC team itself. May I recommend a good vSphere book that will give you the basics of virtual networking?

http://www.amazon.com/Mastering-VMware-vSphere-Scott-Lowe/dp/0470890800/ref=sr_1_1?s=books&ie=UTF8&qid=1317403122&sr=1-1

Good luck!

By: sandip kayande

sandip kayande — Fri, 30 Sep 2011 06:50:14 +0000

Re: Is it possible to give multiple IP address to single server teaming port please tell me….. i am waiting for your replay …

By: Binu

Binu — Tue, 27 Sep 2011 22:37:38 +0000

Hi Scott..

thanks a bunch, i was getting mad about this nic teaming between cisco and esxi server…your blog has given a fair idea..

By: slowe

slowe — Tue, 20 Sep 2011 16:28:12 +0000

Eizhak, some switches only support 802.1Q encapsulation, so the “switchport trunk encapsulation” command isn’t supported (it isn’t needed). You can, though, still use port aggregation as long as it is supported by the switch. As far as I know, a Catalyst 2960G would support port aggregation.

By: eizhak

eizhak — Tue, 20 Sep 2011 07:41:50 +0000

can i use EtherChannel for esxi on lare 2 switch , say Catalyst 2960g
i cant use commands like switchport trunk encapsulation dot1q
for use of IP HASH load balancing.

By: slowe

slowe — Thu, 25 Aug 2011 01:53:23 +0000

Abhishek, that is expected behavior. Any traffic flow between a single source and a single destination will never be able to use more than 1 link in the link aggregate. This is a well-known behavior in the networking world, and there is no workaround of which I am aware (other than using multiple source and/or destination IP addresses).

Good luck!

By: abhishek

abhishek — Wed, 24 Aug 2011 14:01:20 +0000

After I configured the Etherchannel trunk on FE ports between the ESX host and the 2960 switch, I ran some load tests between a VM and a laptop. Even after configuring the “port-channel load-balance src-dst-ip” the Etherchannel did not kick in the second link to push the speed beyond 100 Mbps. This was tested bi-directionally multiple times.

So it seems like even though one may have more than one patch cord connecting the ESX host and a switch, the aggregated traffic is limited to the speed of a single link. Is it not possible to get more than 100 Mbps ( on a FE EtherChannel) for a single connection?

By: Greg Franklin

Greg Franklin — Sun, 07 Aug 2011 03:46:06 +0000

Scott,

I seem to have this almost working perfectly. The only thing I notice is when I am on the switch, and try to ping hosts inside ESXi, it fails. I have a basic setup, no firewalls blocking ICMP traffic. I can however from a VM on vlan 15 ssh to a server on vlan 20, I can also RDP from a vm on vlan 15 to a physical PC on vlan 10. I dont understand what would block the pings im trying to initiate. Could it be routing?

thanks in advance.

By: Nizam Mohamed

Nizam Mohamed — Thu, 04 Aug 2011 03:39:25 +0000

Thanks Scott! Your post really helped me to get my ESX network back online after upgrading my network hardware to a 3750g L3 switch. I never knew about the etherchannel load-balance algorithm despite configuring the port-channel group successfully.

Thanks again!

Nizam.

By: X

Thu, 14 Jul 2011 10:17:26 +0000

If you port trunk at a vlan level without the reliability factor, that is inevitably the precursor for a well defined switching model which doesnt use layer 3 and can give good performance.