Cat blade switch 3020 for HP

g.hammond · ‎06-06-2007

We are in the very beginning stages of evaluating HP blade servers for our enterprise. The blade server uses the 3020 switch for uplinks and internal LAN connections. I've looked at the switch configuration and have a pretty good idea of what needs to be done and I/we plan on testing lots of options and configurations. There is very little documtation on the Cisco website on these switches to speak of. Has anyone worked with these switches that can give me an idea of what works or doesn't work or how they may have it configured for their enterprise? Any doc links would also be quite appreciated. Thanks, Greg

cajalat · ‎06-11-2007

We have several of these configured. They are not exactly the same as the p-Class CGESM switch. For example, where the p-Class has hard-wired Gi0/17 and Gi0/17 these give you the option of 0/23 & 0/24 and then you have to manually configure them to cross connect them (i.e. in interface mode use the "media-type internal" command).

One gotcha is if your end-users plan to use 1/2 slot blades vs the full slot blades. In this case the c-Class enclosure will require 4 3020's and not the typical two in a redundant configuration.

You also have the FastEthernet0 interface which is accessible only via the chassis' iLo port. It is convenient we're finding as it gives us a Side-Band management option.

If there is anything more specific that you're interested in please let me know.

g.hammond · ‎06-14-2007

Thanks for the info.

Obviously, there are several things that our server guys will have to figure out and configure from a server standpoint. I do have a couple of questions, though, that maybe you can help with.

In our current (non blade) environment we have our servers connected to 2 separate core switches set up for fault tolerance plus a 3rd connection on each server that goes to a standalone switch that is connected to a non routed backup network. Would we need a 3rd blade switch to to accommodate this?

Thanks,

Greg

cajalat · ‎08-06-2007

Just realized that I never responded to this. Hopefully you found your answer by now but just in case...

We have the same exact setup for our non-blade environment. For our p-Class and c-Class what we do is have a 3rd uplink from one of the enclosure switches that connects to our backup infrastructure. In our p-Class we generally run one switch as primary and the other switch as secondary (we assign the root bridge on the primary uplink switch). So our backup connection uses the other blade switch for backups. From the standpoint of the server it has two connections that are teamed on each bladeswitch and a 3rd connection for backup on the secondary bladeswitch which leaves one nic unused.

On a c-Class it is identical if you use full blades. If you use the 1/2 blades then you'll have 4 switches per enclosure which you'd connect in a similar fashion.

Casey