[c-nsp] Strange Issue with 3560X and 4500X

[Antonio Soares]

Group,

 

This is one of the most weird things I saw these last years. Imagine a
network where you have two 4500-X in the Core (no VSS) and a few 3560-X
pairs forming squares between the 4500-Xs and the 3560-Xs.

 

One of the 4500-X is the STP root for all Vlans, the other 4500-X is the
backup STP root for all Vlans as well.

 

Between the 4500-Xs and the 3560-Xs I have LACP, CDP and UDLD  running.

 

The issue:

 

The network was up and running well the first 4 days after installation.

 

More or less on the fifth day, all the 3560-Xs connected to the secondary
4500-X, stopped responding to ping requests from anywhere in the network,
even from the directly attached neighbors, the two 4500-Xs and the other
3560-X. A reboot to the 3560-X didn’t  solve the problem. UDLD, CDP and LACP
didn’t fail at all.

 

In order to get normal access to the 3560-X, I had to shutdown the uplink
from the 3560-X to the 4500-X.

 

I have a simple diagram here:

 

http://ccie18473.net/issue-sw2.jpg

 

What seems to happen is that broadcasts (ARP, DHCP) and multicast start to
fail somewhere in time.

 

It must be a very severe 4500X or 3560X bug  but I wasn’t able to find
anything. The most important information:

 

WS-C4500X-32, cat4500e-universalk9.SPA.03.05.00.E.152-1.E.bin

 

WS-C3560X-48P, c3560e-universalk9-mz.150-2.SE.bin, the uplink is fiber
optic, the C3KX-NM-10G is used, between the 3560Xs I have copper

 

Unfortunately I can’t reload/upgrade the 4500X-s or the 3560X-s


 

Any pointers are more than welcome.

 

 

 

Thanks.

 

Regards,

 

Antonio Soares, CCIE #18473 (RS/SP)
amsoares at netcabo.pt

http://www.ccie18473.net <http://www.ccie18473.net/>