[rbak-nsp] PPP timeoute

[Ian Calderbank]

This bears a bit more explanation/discussion:

PPP keepalives are handled by the PPAs on the line cards, so it doesn't make
the XCRP work harder to set them low per se.

HOWEVER if you set the keepalives very low the control plane of the system
will respond much more aggressively to loss of network connectivity to  l2tp
peer(s), (because that implicitly drops all ppp layer traffic)  and thus
boot the subscribers off more quickly. If the box has to boot a lot of
subscribers in a hurry, and then reconnect them very shortly afterwards
(because your route to the l2tp peer came back), this makes the RP work
really hard. It's perfectly possible to melt your SE that way if you have
enough subscribers, particularly on XCRP3. It will sit at 99% on the RP and
never recover to stable state - as the subs try to reconnect, it doesn't
have enough spare cpu to fully connect them (aaad, l2tpd, ism are all
working like dogs)  the subs get part way through reconnection then drop
again .. ad infinitum. You need to get into understanding and tweaking the
throttling timers  to deal with this. 

So setting the ppp keepalive too low could be a serious cause of XCRP load.
I wouldn't suggest to down to 10 seconds x 3  keepalives for PPPoL2TP subs,
if it was a big  box that could make the RP melt.  I'd be thinking about if
it was possible to tune the routing instead.

cheers
Ian

------------------------------

Message: 4
Date: Wed, 12 Jan 2011 13:25:32 +0100
From: Marcin Kuczera <marcin at leon.pl>
To: sledge121 at gmail.com
Cc: redback-nsp at puck.nether.net
Subject: Re: [rbak-nsp] PPP timeoute
Message-ID: <4D2D9DBC.6080705 at leon.pl>
Content-Type: text/plain; charset=ISO-8859-1; format=flowed

Richard Clayton wrote:
> Does anybody know the default PPP timeout and the command to adjust 
> it., I have 2 Ethernet circuits carrying PPP over L2TP, the circuits 
> are configured as active / standby using BGP metrics, the BGP hold 
> time is
> 30 seconds and I was wondering if the ppp users would disconnect 
> before BGP flipped the routing over to the other circuit.

  ppp keepalive check-interval seconds 10

be carefull about XCRP load.

Regards,
Marcin


------------------------------

_______________________________________________
redback-nsp mailing list
redback-nsp at puck.nether.net
https://puck.nether.net/mailman/listinfo/redback-nsp


End of redback-nsp Digest, Vol 37, Issue 8
******************************************