[c-nsp] GSR: 3GE-GBIC-SC v4 traffic influenced by v6 traffic?

[Jan Sandmaier]

Hi Sascha,


> the 3GE-GBIC-SC card does IPv6 in software on the linecard. Is anyone
> aware of a problem that IPv6 traffic that is software-forwarded
> could influence IPv4 hardware forwarding? It looks like a linecard
> could hit 100% with ~80 Mbit/s of v6 traffic + other tasks
> like TAG Stats Backgr or CEF process etc and also suddenly stops
> forwarding IPv4 traffic or AT LEAST stops responding to ICMP Echo
> (directed to the interface IP) or loses IP protocols like
> LDP or OSPF which could point to problems GRP/PRP -> Interface.
> It seems like this happens exactly at that moment when the
> card's CPU hits 100%.

this was exactly the reason why we de-activated IPv6 on these lincards. 
This applies for all Engine 2 linecards.

>
> I had expected that the CPU forwarded v6 could lead to higher
> RTT for v6 etc but I am surprised that it seems like it could
> influence IP protocols.
>
> I know that the Engine 2 card is not state-of-the-art especially
> for v6 traffic but I am looking for a confirmation that this
> behaviour is indeed something that could happen. If so, I might
> go to for 4GE-SFP-LC or similar which does v6 in hardware AFAIK.

4GE-SFP-LC and any other Engine 3 linecard works well..

Best regards,
Jan