Configure 2 ProCurve 2510G for VM ESXi iSCSI

[smoothrivers]

I am building a VM datacenter consisting of 6 VM hosts connected to 3 MSA storage devices.  I have 2 procurve 2510G switches to use to connect the hosts to the MSAs.

I have port 24 on both 2510s configured as the management port, connected to my regular data network.

I have ports 1-18 on both 2510s in an "untagged" VLAN called VMiSCSI.

Right now there is only 1 2510 attached to the VM datacenter, and am wanting to configure the 2nd 2510 and install it.

Both switches will remain segmented from the rest of the network.

Is there anything special that I need to to do to get the second 2510 to talk on the VMiSCSI VLAN, or can I just plug them together via port 23?  Should port 23 on both switches also be in the VMiSCSI VLAN?

Should I take ports 20-21 on both switches and build a "trunk" between the two, and if I should, how do I do it?

Sorry if the questions seem silly, but I am not a network guru, and thanks for the help.

[John Gelten]

For the connection to your data network :

Keep in mind you will probably be creating a loop - at least on the physical level, so some kind of spanning-tree protection might be advisable; obviously one that is compatibe with your existing network. And please DO change the default priority of your new switches to a lower-priority value or you could make them the root of your entire network (I am hoping your data-network has been set up in a way that this is not possible, though). Another way to prevent a logical loop is to give both switches a single VLAN towards the data-network that is independant from the other (kind of a 'routed interface').

For the connection between the two switches :

Depending on the topology of your SAN, you may or may no expect a lot of traffic over that interlink.

If the targets and initiators are consistantly placed on the same switch, there will not be a lot of traffic under normal circumstances. But when targets and initiators are randomly connected to both switches, you will statistaically end up with 50% of the traffic leaving every host-interface needing to traverse to the other switch.

In that case you may want to create some extra bandwidth between the switches by trunking two or four interfaces.