[f-nsp] MLXe interface management - one mac address per chassis ?!?

[Youssef Bengelloun-Zahr]

Hello,



I have stumbled upon a limitation that I can't seem to find any workaround
for. I have the following setup per-pop :

MLXe ========== FCS stack (2 members) =========== Management interfaces of
devices (switchs, PDUs, firewalls, etc.)

I created a managment VRF that spans multiple POPs, each has it's IP
private space.

A Vlan (on the MLXe) is binded to a VRF, the VLAN is trunked down to a
stack of switchs (using a LAG) in order to aggregate cabling.



Problem I have is the following :

I can't use the dedicated managment interface on the MLX MR2 modules
because the chassis only has one mac-address.

So, when I plug in the cable to the aggregation switchs, mac-address is not
learned from that port but rather from the LAG. Otherwise, I would create a
nice L2 loop ;-)



This on the MLX :

Ethernetmgmt1 is down, line protocol is down

  STP Root Guard is disabled, STP BPDU Guard is disabled

  Hardware is Ethernet, address is *0024.38a5.7b00 (bia 0024.38a5.7b00)*



This is on the FCS stack switchs :

telnet at ag01-par01#sh mac-address

Total active entries from all ports = 35

MAC-Address     Port           Type          Index  VLAN

....

*0024.38a4.fb00  1/1/1*2/1/1    Dynamic       28312  100*



N.B : e1/1/1 and 2/1/1 are the ports used to create uplink LAG to the MLX
router / VLAN100 is the managment VLAN bounded to the management VRF.



For operational needs, I would really like to use the dedicated management
interface on the MLX routers (authentication, supervision/monitoring, etc.).

Sadly, there is nothing you can do on that interface except change the IP
address :-/

Has anyone faced this before ? Any workaround you can think of ?

Best.

Y.
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <https://puck.nether.net/pipermail/foundry-nsp/attachments/20130130/ce9bbe29/attachment.html>