recomended networking setup with 4 network ports (link aggregation)

[rakem]

Hello, I have a few questions about configuring ESX servers that have 4 network ports.

We have 5 esx servers running in our environment, each server having 4 network ports. (Two dual ports nics)

There is:

vmnic0 and vmnic1 on one card

vmnic2 and vmnic3 on the other card

In VMWare these are teamed vertically so:

vmnic0 is teamed with vmnic2

vmnic1 is teamed with vmnic3

This is the way that was recomended to us by the guys who helped setup our servers, is this the best way to configure the nics in VMWare?

Now, currently the switch ports that these cards connect to are not aggregated. We have two Dell switches in a stack which will be used to setup LAGs for all the ESX servers.

The way that i think this should be setup is:

LAG1 on Switch 1 for the vmnic0 and vmnic2 team

LAG2 on Switch 2 for the vmnic1 and vmnic3 team

So each team is aggregated and connected to a different switch.

Does this sound like feasible configuration that will result in proper redundancy?

Thanks!

[depping]

Yeah this is the way to go! Keep in mind if you're loadbalancing the physical switch port than the best thing to do is set the vswitch to loadbalancing on ip. do manual loadbalancing for the physical switches, no automatic stuff...

Duncan

Blogging: http://www.yellow-bricks.com

If you find this information useful, please award points for "correct" or "helpful".

[rakem]

Thanks for the response.

Got a very very weird issue now. I have setup everything as per my first post however, as soon as i connect the second nic into its lag group my pings start timing out.

Example - I was working on the LAG for the service console which is 10.1.21.7. I added vmnic2 into the LAG and plugged it into the appropriate switch port. No problems, pings continue.

as soon as i added vmnic0 into the switch port in the LAG my pings dropped out.

I thought it may be a spanning-tree issue but i have disabled all spanning-tree options on those ports and the switch reports that all ports in the LAG are up and have correctly negoiated the correct speed and duplex.

If i unplug any port in the lag the pings return, but again as soon as i plug in a port pings fail...

EDIT - Just a thought. Im using Dell switches and for the LAGs i was just setting them to be hard coded as on. I have now changed them to use LACP and set them to auto. When they are auto i have noticed that the LAG remains inactive.

So i believe the problem may lie with the vswitches not being able to recognise the LAG. Is the something that needs to be done so i can either user LACP on the vswitch or even just hard code them to be on without any negoitation??

Any suggestions?

[rakem]

ok i have solved the issue.

Turns out that you need to select 'Route based on IP Hash" in the load balancing for the vswitch.

[depping]

That's what I suggested in my post, use IP for load balancing... 🙂

Duncan

Blogging: http://www.yellow-bricks.com

If you find this information useful, please award points for "correct" or "helpful".