Dell Blade M1000e and Trunking

[sansaran]

We have dell Blade M1000e enclosure with 5 full length and 5 half lengh blades. we gonna be installing ESX5.0 on full length blades.  I have a question on ports which are need to be configured as trunk ports

All blade servers connected to redundant Internal ciso switches(3130s).  we have 4 internal cisco ethernet switches installed (3130s) installed on A1,A2,B1,B2  i/o modules of the M1000e.   All ethernet switches are stacked up together by network team.

Each Internal switch has 16 internal ports(downlink ports) and four external ports( port 17,18,19,20).

Each Full length blade has 8 nics and all connected to downlink port of Internal switches as follows

Switch                connected ports

  A1              1,2,3,4,5,9,10,11,12,13

  A2              1,2,3,4,5,9,10,11,12,13

  B1              1,2,3,4,5,9,10,11,12,13

  B2              1,2,3,4,5,9,10,11,12,13

External port 17 on each cisco internal ethernet switch connected to external network switch.

we have couple of Vlans configured on both external and internal swicthes. lests say  101 and 102

My question really here is ,  which ports should be configured as trunk ports

(i) All internal (40 ports) ports in the cisco blade switches and external port (17) which is connected to external switch

(OR)

(ii) Only external port 17 on each cisco blade switch should be confugures as trunk port

[Dhanushka Haputhanthri]

Hey, did you ever figure this out? I have a the same question, anyhelp you can provide would be great.

Thanks.

[ABCRacer]

Hello Sansaran,

I am very familiar with this configuration.  A better question fielded back to you is why not configure the 10Gb uplinks to be the trunk ports on the 3130s?  If that's not the case, create a port-channel from whatever copper ports are available to uplink the trunked 101,102 VLANs to provide redundancy and bandwidth for all your blade ports.  Let me know if that helps or not.

Jon

[nilesh56]

agreed with ABCRacer

Trunking all internal (40 ports) ports in the cisco blade switches and external port (17) which is connected to external switch.

Also configuring LACP over 8 internal ports of each hosts and 4 external port.

Let all VLANs configuration be within your Portgroups in VMware environment and over your physical switch to avoid complexity and management overhead.

[HeathReynolds]

In most ESX configurations your upstream links (port 17 in your case) will always be trunks, and all of the internal connections facing blades running ESX will be trunks as well. This will allow you to run multiple VLANs for VMotion, Management, NFS, and guests. You have enough interfaces avalaible on each ESX host that you could just run a single VLAN on each, but having each interface configured as a trunk carrying multipe VLANs a much more common configuration.

An example config for one of the internal ports facing an ESX host would be :

(config)# interface GigabitEthernetxx/xx

(config-if)# description <text>

(config-if)# no ip address

(config-if)# switchport

(config-if)# switchport trunk encapsulation dot1q

(config-if)# switchport trunk native vlan <vlan id> (optional, I don't run it but it makes ESX easier to bring up)

(config-if)# switchport trunk allowed vlan <vlan id or list>

(config-if)# switchport mode trunk

(config-if)# spanning-tree guard root

(config-if)# spanning-tree portfast trunk

Your interfaces going to the half height non-esx blades could go either way, but in most enviroments would probably be access ports carrying only one vlan.

Here is a good .pdf on the 3130 that may help you :

http://www.cisco.com/en/US/prod/collateral/switches/ps6746/ps8742/ps8764/white_paper_c07-443792.pdf

You have some options such as LACP port channels spanning all four 3130s if they are in the same VSB domain. This leads to some interesting options for how you build networking inside of ESX.

Will you have the vDS or 1000v avaliable, or are you using the standard switch?

[mydearcosmo]

Hey, guys. Any update on this? I also have the similar case and just waiting for an answer. Thanks.