HP Virtual Connect Mac-Address Flap

[bishwa]

Hi There,

I have HP C-Class chassis with 6 blade servers. The servers connect to external Cisco 3750 switches through HP Virtual connect modules.The servers are on the same vlan.
The Cisco 3750 switch ports are configured as access ports with spanning tree portfast.

There are 4 Virtual connect modules. Bay 1 and Bay 2 are connected via the backane and so are Bays 3 & 4.
Then we have external stacking cables connected between Bay 1 & 3 and Bay 2 & 4.

The Virtual connect modules are configured as follows:

1/ 1 Server Profile per server
2/ 1 network profile (uplink) per server.

The issue I am having is that the Cisco switches are reporting MAC address FLAP on ports that are connected to the same Virtual connect. For example, Cisco switch ports 2/0/1 & 2/0/2 connects to Virtual Connect module 1 ports 1 & 2 respectively, then ports 2/0/1 & 2/0/2 will complain about mac-flap.

The same will happen to on VC module 2 or 3 or 4.
The mac addresses that the Cisco switches complain about are the Mac addresses of the Virtual Connect Stack Interfaces.
I have got around the issue by configuring the all server profiles to go via one network profile, which effectivly means one link.
There is the option of LACP & DOT1q tagging but the customer does not want to use that at this stage and want the reason for the current issue.

I think that the VC is sending LLDP with the Stack interfaces Mac address, on all Network Profiles. This causes the same MAC address to appear on multiple ports, as in my case, on the Cisco switches which then complains about MAC-FLAP.

Has anyone out there had this problem and if there is fix for it.

Thank you,

[Lee Carey]

Hi, Did you manage to resolve this. Having a similar issue

Thanks

[bishwa]

Hi Lee,

We did a work around by trunking all the devices through the 10Gig Trunks.
However, some of our engineers went on VC training course and they asked the question as to why we are having the issue. They were advised by the instructor that its due to an issue in the firmware of the Virtual connect module.
We have not persued this any further as in the end we connected all the servers via the 10Gig trunks.
I would suggest asking HP and see if they can help.
I would appreciate if could let us know what the findings were from HP, if you do end up talking to them.

Sorry I could not be anymore help.

[cls.co.za]

Look at

http://bizsupport.austin.hp.com/bc/docs/support/SupportManual/c01471917/c01471917.pdf
Scenario 9 & 10

From my understanding you need to plug both VC's on the Left i.e. interconnect 1 & 3 to the same switch and the ones on the right i.e. interconnect 2 & 4 to the same switch. When you configure your O.S. with a team/bond and use the performance/automatic option and not the fail over option you will see this flapping (if you have configured your cabling as you described) you will have the same server requesting traffic on 2 different MAC addresses hence the "flapping". You would only be able to achieve this if your switches supported "distributed trunking"

SO basically left interconnects to core 1 and right interconnects to core 2

The stacking cable between interconnect 1 & 3 and interconnect 2 & 4 would also be recommended.

Hope that this helps