01-08-2013 10:54 PM - edited 03-04-2019 06:38 PM
Hi
Yesterday afternoon we experienced an issue where 2 Cisco 7609 (Sup 720 3BXL) routers functioning as PE devices maxed out the CPU for a lengthy period of time. This caused huge BGP instability. The main processes taking the CPU time were either the BGP scanner which I'd expect and LFDp Input Proc.
I can't seem to find a lot of information about this process. Could anyone provide some more detail as to what this process does? The end result to 'fix' the issues was to remove the majority of routes being received by the 7609 turning it more or less into a P device. We do have several other 7609's running with a full BGP table in a identical topology that haven't shown this issue.
Thanks
02-02-2014 08:00 PM
Today, I seen this problem. Why doesn't who answered
03-12-2015 12:42 AM
Hello, everybody.
Sure it's not the best time to answer, but I hope this could help others.
The process "LFDp Input Proc" is responsible for software switching of MPLS packets (like "IP input" for non-tagged packets).
So, in your case the problem is - a lot of MPLS packets are hitting CPU. This might be caused by a couple of reasons - bug (hardware programming), LFIB exhaustion, configuration issue, any kind of flaw in multicast, TTL le 1 and etc.
Troubleshooting for LFIB issue - show mls cef adj usage; others - to capture traffic being punted to CPU - with netdr or debug mpls packet process/internal.
Hardware mis-programming COULD be fixed with interface shut/no shut or routing flap.
Discover and save your favorite ideas. Come back to expert answers, step-by-step guides, recent topics, and more.
New here? Get started with these tips. How to use Community New member guide