IBM BladeCentre and Cat4506 - Port Flapping

[gprokschg]

Hi all,

we recently got some IBM blade centres and certain staff have just been patching the Ethernet ports from the back of the blade centre to various ports on our redundant 4506 cat's. Now I have heaps of messages saying that various VLAN's are flapping. I know this has only started to happen since the blades were put in. All ports that have servers connected have "spanning-tree portfast" enabled. I think I have to turn this off as the blade chassis seems to use the same MAC address for various blades inside it.

On top of this, some of the blades are going to be loadbalanced.

Has any got any set up info for this (experienced the same issue etc)?

The only document I found on Cisco was a IBM Bladecentre design guide, which didn't seem to have a lot of tech info in it.

Cheers,

G

[skarundi]

if you are talking about a message that says something like this "/Host 00:50:0f:20:08:00 is flapping between port 1/2 and port 4/39" this is normal in your case. The cat4k supervisors produce a message when it learns a Unicast MAC address from 2 different ports on the same LAN within a few seconds.

Use a multicast Mac address for the virtual IP of the blade centre. that is MAc address that start with 01, 03, 05, 07..(basically the 2nd hex number is odd ). Then manually assign this multicast MAC address to appropriate switch ports. After that on the default gateway ( if the router is a Cisco router ), manually assign an ARP entry for the virtual IP, binding it to the multicast MAC address.

This should take care of the messages.

Also, you can just ignore the messages if all is working well.

Normally the host flapping message is a problem only when you have a bridging loop.

[paddyxdoyle · Accepted solution]

Hi,

IBM has 2 good documents that show how to configure the ESM blade switches in a Cisco environment (the ESM switches are manufactured by Cisco!).

http://publib-b.boulder.ibm.com/Redbooks.nsf/RedbookAbstracts/redp3660.html?Open

http://publib-b.boulder.ibm.com/Redbooks.nsf/RedpieceAbstracts/redp3869.html?Open

I have just done similar to you, connecting the blade switches to 4506's and 6509's but without loadbalancing. The only issue we had was realising that each blade has two network interfaces, these interfaces are hard wired to individual switches, so for example, if you lost a switch1 which is hardwired to LAN1 on the blade and didn't have network teaming configured, you would lose connectivity to your server. If you had teaming configured then the team will start using LAN2 and thus ESM switch2. FYI to use teaming you need to install the BASP drivers.

We didn't get any problems with ports flapping, at one stage the ESM became the root switch for our production VLAN so i had to add port costs to all the blade ports, incidently, IBM recommend turning off spanning-tree on the Blade switches which i will hopefully be able to test next weekend.

Hope this helps

Rgds

Paddy

[gprokschg]

Thanks guys. I have read and understood the IBM RedPaper. Will implement trunking between blades and Cat 4500.

Cheers,

G