[c-nsp] Multilink PPP high reordered packets and lost packets

[Rodney Dunn]

On Fri, Oct 21, 2005 at 09:03:25AM -0400, Douglas E. Warner wrote:
> After getting my MLPPP bundles to come up finally, now I have a new set of 
> problems that seems to affecting performance.
> 
> This looks to be a problem similar to this thread:
> https://puck.nether.net/pipermail/cisco-nsp/2004-July/011529.html
> 
> I've got MLPPP setup between a 7505 and a 7507 running RSP4s on 12.4(3) over 
> 4xT1 on PA-MC-8T1 cards in VIP2-50s running dCEF.  I've set 'ppp multilink 
> fragment disable' on both ends of the bundle, but I'm still seeing the 
> following:

Turning that off is fine but it's irrelevant as we don't do any fragmenation
on the 75xx unless you have LFI configured.

> 
> Multilink1, bundle name is rtr0.tc
>   Endpoint discriminator is rtr0.tc
>   Bundle up for 1d08h, total bandwidth 6144, load 53/255
>   Receive buffer limit 48000 bytes, frag timeout 1000 ms
>   Bundle is Distributed

so it's distributed.

>     42/0 fragments/bytes in reassembly list
>     50191 lost fragments, 24370398 reordered
>     0/0 discarded fragments/bytes, 0 lost received
>     0xDEBE4E received sequence, 0x2E8D53 sent sequence
>   Member links: 4 active, 0 inactive (max not set, min not set)
>     Se4/0/0:0, since 22:05:14
>     Se4/0/1:0, since 22:04:57
>     Se4/0/2:0, since 22:04:44
>     Se4/0/3:0, since 22:04:33
> No inactive multilink interfaces
> 
> Similar results on the other end.
> 
> Doing the command Rodney suggested in July 2004,
> # show controllers cbus | inc Serial4/0/0:0
>       Serial4/0/0:0, txq E8001A00, txacc E8000102 (value 25), txlimit 26

And the master of the mlppp bundle has plent of txacc's available
to transmit so you don't have congestion.

> 
> although, over time these values don't seem to be changing.
> 
> Any thoughts on what I can do to debug this?

You could shut down one member link at the time and see if one
is causing the problem. 

Are you seeing the same thing on the other side?

The other option is a multilink debug on the vip and manually go through
and compare all the sequence numbers and see which ones are dropped
or coming in out of order. That's not an easy exercise.


a> 
> -Doug
> 
> -- 
> Douglas E. Warner    <dwarner at ctinetworks.com>     Network Engineer
> CTI Networks, Inc.   http://www.ctinetworks.com    +1 717 975 9000



> _______________________________________________
> cisco-nsp mailing list  cisco-nsp at puck.nether.net
> https://puck.nether.net/mailman/listinfo/cisco-nsp
> archive at http://puck.nether.net/pipermail/cisco-nsp/