[f-nsp] NI40G interfaces blocking

Henk.Roose at cwi.nl Henk.Roose at cwi.nl
Fri Aug 3 07:40:56 EDT 2007


Henk Roose wrote:

> Charley Kline wrote:
> 
> > Line card problems on the 40G are nightmares to diagnose. We had a  
> > similar problem on a couple of ours that they worked on for months,  
> > complete with special instrumentation on the LP's and an engineer on- 
> > site. They finally isolated it to too much broadcast traffic (ARPs,  
> > due to a network loop somewhere) overwhelming the LP, which then  
> > stopped responding.
> > 
> > Things seem much better at 2.3.0d, for what it's worth.
> 
> They (foundry) suggested an upgrade to 2.3.0d, which I did last
> monday.

A lot better indeed. The box has been running without any problems
for three and a half weeks.

> I'd like to add that the problem seems packet processor specific.
> The same lock-up behaviour occured on a 40x1G port line card. That
> card has 4 PPCR's and the 10 ports associated with PPCR 1 where dead.
> All others were still forwarding traffic.
> 

Here's the bugreport:

-->8--
Bug ID # 70986
Description:
Seeing buffer loss IPv6 testing.  The buffer loss may cause a transmit
'lockup' behavior.

This bug is in relationship to IPV6 concerning buffer loss occurring
with running IPv6. Pending IPv6 packets queued creating Loss of buffers
due to pending of packets on deleted ND6 entry. Because of this behavior
it's causing buffer loss. This may explain transmit 'lockup' behavior.
This is why we are recommending upgrading the customer to 2.3.00d
-->8--

Henk

-- 
Henk Roose
CWI - Centrum voor Wiskunde en Informatica
Centre for Mathematics and Computer Science
Amsterdam (NL)



More information about the foundry-nsp mailing list