01-17-2013 01:10 PM
We have two MLXs connected via a trunk.
We are stable on 04001a but upon upgrading to 05400b we are seeing crippling high CPU utilization on our MPs and LPs.
This was in one of the MLX logs:
alert S CAM IP VPN partition warning: total 327424 (reserved 49152), free 16371, slot 1, ppcr 0
Our CAM usage is 10% of that on 4 code.
We can't help but think that this is related somehow to multicast because we have a large amount of routed multicast traffic on our network. (We are running PIM-SM in multiple VRFs. In addition, show task reveals that mcast process is elevated when things break.
I'm thinking someone else must have run into a similar issue upon upgrading from 4 to 5 code. We're not getting a lot of responsiveness from TAC...
Also, this is in the diagnostic guide which seems suspicious:
This can occur when the device runs out of CAM due to large amounts of multicast traffic flooding the CAM. Make sure the device is using the most current patch of the software, and contact Brocade Technical Support for assistance.
We are using this CAM profile: mpls-l3vpn-2
I'm hoping there was a default configuration change that we need to account for when going from 4 to 5 code that is causing this behaviour, specifically with respect to how the device forwards multicast in hardware.