[c-nsp] per-LSP packet loss / FIB corruption?

[Phil Mayers]

Saku Ytti wrote:
> On (2009-07-16 10:23 +0100), Phil Mayers wrote:
> 
> Hey,
> 
>> Could it be the slot? If so, why would it manifest only on a single,
>> or a small number of LSPs?
> 
> I've had various of packet loss issues  affecting just single prefix
> in PFC3x boxes, typical cause is invalid programming in hardware,
> but correct in software, causing hardware to punt traffic
> rate-limited to software and software forwarding correctly.

Ah; and we've got quite aggressive CoPP and punt rate limiters. Would 
that mean the traffic would be dropped quite aggressively as it was punted?

> 
> Best way to debug when you've eliminated config errors and
> physical link issues is to use ELAM to capture DBUS/RBUS
> headers, which will tell you, what the platform is going
> to do to the frame.

Interesting; ELAM is not something I've ever used before. I see there's 
a doc on Cluepon - I'll have to take a look.

> If interrupt load is something different than 0-1, most
> likely something is wrong, however punted amount of 
> traffic may be so low that something is wrong even though
> interrupt load is 0-1.
> 

Interesting.

As I said, in this case we rebooted the linecard and the problem seems 
to have gone. Are there other routes that will reliably reprogram the 
hardware? The "mls cef" inconsistency checker seemed to think all was well.

Thanks very much for the suggestion; it fits the symptoms extremely well.