Blade Chassis network design configuration

[Addisonfun]

What would be best practice for a new vmware environment that is a HP blade chassis with 4 3020 cisco switch and 2 brocade switch.

The blades are half blades so it would be a total of 4 physical nics.  I was planning to port channel 4 fiber ethernet ports per switch.  All 3020 switches will feed back to a single core switch.  Each brocade will each feed to two san switches.

[a_p_]

Welcome to the Community,

sounds like a design which should work just fine. What I would recommend - that's what I usually do - is to configure link state tracking on the Cisco switches. This way the ESXi's vSwitches will properly fail over in case a physical switch looses all uplinks.

André

[Addisonfun]

I guess I am seeking more information about configuration about the NIC configuration in the Vsphere cluster.  How would you configure the 4 NICs, How or which ones would you team.  This is with the normal Vsphere Enterprise license.

[a_p_]

Depends on the features you are planning to use. For a default installation with 4 NICs (Management, vMotion/DRS, VM traffic) I'd probably use two uplinks (one to each physical switch) for Management and vMotion and the other two uplinks for Virtual Machine traffic. vSwitches configured with the default policy "Route based on originating port ID". All uplinks configured as trunk (802.1Q) ports.

as an example:

vSwitch0 - vmnic0, vmnic2 (both active)

--> Management Port Group - Management VLAN with vmnic0 active and vmnic2 standby

--> vMotion Port Group - vMotion VLAN with vmnic0 standby and vmnic2 active

vSwitch1 - vmnic1, vmnic3 (both active)

--> Virtual Machine Port Groups with appropriate VLANs configured and all vmnics active

This may be different in case you are going to use e.g. FT which requires it's own vmnic.

André

[Addisonfun]

Okay, this is what I was kind of thinking but still seems like a  waste of uplink ports for just management and vmotion.  Thanks for the  input!

[a_p_]

It might look like a waste when you first look at it, but if you consider that vMotion should have it's own uplink and the Management should be redundant you see that both requirements can be achieved with a configuration like this. You certainly won't need vMotion all the time, but this is actually the same as with a car. You only need it for driving, but most of the time it's in the garage or on a parking lot (what a waste :smileywink:).

André

[Addisonfun]

lets go back to the physicl network now  :smileyhappy:   So on each blade switch I configure the port-channel port to have say "link state group 1 upstream" and each blade port to have "link state group 1 downstream"   ..

How do I configure the core switch?

[a_p_]

Yes, that's what you will do on the blade switches to ensure all downlink ports go down if all uplinks fail.

On the core switch you have to configure the channel as usual, there's nothing special that has to be done due to the Link State group on the blade switches. If the core switch has multiple modules/slots, connect the uplinks in a way that either a blade switch or module can fail without interrupting production.

André