Management and VMs, place in different vlans?

[HendersonD]

Please refer to the diagram below
We have two 1GB switches in our Bladecenter that are just used for the vSphere Management network

The management network is in VLAN 125. For example, on the attached diagram, the host shown has a Management IP of 10.121.125.161

We have two 10GB switches in our Bladecenter that are used for VM, Storage, and vMotion traffic

The VMs (servers) are also in VLAN 125. For example my Exchange mailbox server has an IP address of 10.121.125.80

I think having my Management and VM network in the same subnet is causing some spanning tree issues

We had a topology change last week after some switch work and both physical connections to the 1GB switches went down and I suspect spanning tree downed them

I am thinking about moving my management network to a different subnet all by itself thus eliminating two different paths for VLAN 125

Thoughts?

[Josh26]

There's two different questions here.

As soon as you say "management and VM", people are going to chime in saying "of course these should be separate VLAN as per best practice".

However, if you have some sort of network loop triggering spanning tree - this isn't the way to fix it. Nothing involving spanning tree should cause a switch to shutdown both ports to an ESXi host.

[HendersonD]

The spanning-tree piece is exactly what I thought but I have no ready explanation why both of my switches that hook to the Management network had their ports disabled. If you look at the attached diagram you will see my setup. I have 4 connections to the two 10GB switches, they are setup as an LACP aggregate. Two of the connections were showing as not participating in the aggregate, these are noted in red. I simply removed them from the aggregate and added them back in. This triggered a spanning-tree recalc (which it should) and my two 1GG connections to the Cisco switches went down.

I will be moving my management network to a new VLAN

[Josh26]

If you are using the Cisco switches in a bladecenter, by default, they have an internal uplink cable. Depending on where your root bridge is, you could knock one of those paths off as less preferred. Shouldn't affect both though.