Stacking

[lisamiller0019]

I'm new to Procurve switches and also to the concept of stacking switches. I'm also new at my current company so I have inherrited 4 Procurve switches that were already in a stack. The commander is a 2610 and its members are 2601's and one 2810. We're putting in a VMware environment with a SAN along with 2 new 2810 switches. I added the 2 switches to the existing stack and then created trunks across the 2 switches.

My first question is: Do you have to stack the switches before you can trunk ports across them? (i.e. is one a feature enabled by the other)

But my bigger problem is something very strange. The first 2 ESX servers I configured are working perfectly with trunks and VLANs across the 2 switches. The third however has a strange problem. The controllers on my SAN will not talk to any odd numbered IP addresses but they will talk to even numbered IP addresses. E.g I have 2 virtual NICs on the same VLAN as my SAN controllers and I cannot ping the odd numbered one but I can pinf the even numbered one. If I swap the cables over on the NICs my controller can then ping the even number IP but not the odd.

The third ESX server was created using a VMware profile so the virtual networking is exaclty the same as the 1st and 2nd. I've tripple checked the switch config (trunking, LAN, physical cabelling) with a colleague and it all looks the same there too. The only think I can think of is that the Commander is a different model than the 2 new switches so I am thinking of splitting them out in to their own stack.

Can anyone confirm whether I will loose my trunk & VLAN config by removing the switches from one stack and then adding them to another?

[Mohammed Faiz]

Hi,

I haven't played with stacking much but I don't believe that stacking allows you to run trunks across seperate switches.
The only feature I've seen for this is the Distributed Trunking available on the 3500/5400/8200 series code.
What kind of trunks do you have setup (LACP or static)?
The ping issue that you describe sounds like normal behaviour for a broken trunk. Traffic load-balancing is based on source address/destination address pairs. So if one of the links in your trunk isn't working (or on end of the trunk doesn't see the links as a trunk) then some of your traffic will be fine the rest will get "lost".
Removing a switch from a stack shouldn't affect the configuration on the switch.

[lisamiller0019]

Thanks for the reply Mohammed. All of my other trunks across the 2 switches are working fine - I have tested this by pulling the plug on one of the NICs :-) But your description of a broken trunk does make sense. I am just using static trunks.

[Mohammed Faiz]

What does your trunk config look like? i.e. How many nics have you got on your ESX server, which switches do they connect to and what's the relevant trunk config on the switch for those ports?

[lisamiller0019]

I have 6 NICs in each ESX server. There are 3 VLANs with 2 NICs (per server) in each. One of these NICs is plugged in to the first switch and the other in to the second. These ports and then assigned the same trunk ID (trunk type = trunk), and the trunk is then tagged for the revelant VLAN.

[Mohammed Faiz]

So you haven't got any trunk configuration on the switches themselves?