Trunking between cat 4006 and IBM Blade Center

rbradfield · ‎03-04-2003

Customer is installing IBM Blade centres, these boxes take up to 14 servers, and have an in built Gigabit copper switch interconnecting the servers and provide connectivity to the LAN, Connecting as a single VLAN to the cat 4006 is no problem, but trying to setup 802.1Q trunking, I could not get a good connection.

has anybody had experience with setting up the IBM Blade Center Switch for 802.1Q trunking to a Cat 4006

Thanks

milan.kulik · ‎03-05-2003

Just one tip:

Have you tried set dot1q-all-tagged command?

IBM might not be using native VLAN (it's a Cisco proprietary implementation).

Regards,

Milan

rbradfield · ‎03-05-2003

Milan,

thanks for the tip, but this is a global command and when my customer tried this command it stopped all the other trunks ( to routers and 3524 switches) from working properly, could not get to anything on the native VLAN

can the cat 4000 take this comand:

set port dot1q-all-tagged 3/24 enable .. to just tag the required port ?

I cannot play with this till the weekend to test

Richard.

mschnabe · ‎07-02-2003

What was the end result?

Mark

rbradfield · ‎07-02-2003

The customer left it as a single VLAN.

So did not get a solution to the trunking problem.

mschnabe · ‎07-03-2003

There is a brand new redbook from IBM on how to set up blade servers/ Blade Center with Cisco switches.

http://www.redbooks.ibm.com/redpieces/pdfs/redp3660.pdf

ablak01 · ‎08-11-2003

I have setup the IBM Blade servers on a 2948G switch. We used port aggregation, so I had to enable port channelling on the 4 ports that plugged into the switch. Also, IBM had us connect the blade server switch with crossover cables and a special connector. You may need to check into that. I have the blade servers on a seperate vlan and cleared all vlans off the trunk ports defined for the blade servers. I also had to set these trunk ports to nonegotiate dot1q, even though all other trunks are set to desirable. Be careful connecting the blade servers up. The first time we tried, we took the network down with the spanning tree flapping that occurred when the blade servers were not configured properly.

mioconnor1 · ‎08-11-2003

We are having similar issues with our bladecenter. Multiple VLANs work like a charm. Native VLAN for managing the switch works good. This all works with a single link up. But when a 2nd link is up between switch module 2 and a 2nd 6513, network traffic works great for a couple of days, then it drops. The only log that shows up is a duplicate address ARP confusion in the MSFC2s of the 6509s. I really would like to speak with you about how you got this to work. Look at page 149 of the IBM Redbook and that is how I am doing this right now. IBM claims to do the scenario from page 151 with LACP port aggregation. This seems to make the issue more complex in my mind.

ablak01 · ‎08-13-2003

We use LACP port aggregation as well. The blade center has 4 ports in each switch that uplink to the 2948Gs I have them connected to. The first switch has 4 ports aggregated and linked to a port channel that is configured on one 2948G, and the second switch is done the same way to a different 2948G. The 2948G stacks are dual homed to the two core 6500s. When we did it any differently, we had the same problem with the mac address flapping between ports, because the switch would see the mac address of from the IBM switch coming in on the port it was in and also through the redundant link to the second 6500.

mioconnor1 · ‎08-11-2003

We are having similar issues with our bladecenter. Multiple VLANs work like a charm. Native VLAN for managing the switch works good. This all works with a single link up. But when a 2nd link is up between switch module 2 and a 2nd 6513, network traffic works great for a couple of days, then it drops. The only log that shows up is a duplicate address ARP confusion in the MSFC2s of the 6509s. I really would like to speak with you about how you got this to work. Look at page 149 of the IBM Redbook and that is how I am doing this right now. IBM claims to do the scenario from page 151 with LACP port aggregation. This seems to make the issue more complex in my mind.