[rbak-nsp] L2TP extended failover

[Bernd]

Hi list,

we're happily running a bunch of Linux machines using l2tpns terminating 
several tens of thousands of L2TP sessions. Until a year ago, without 
any problems, for almost half a decade.

A year ago our upstream provider changed its BRAS from some Cisco gear 
(IIRC) to Redback -- from then on, L2TP extended failover stopped 
working.

We reworked the code so that we can configure every valid combination of 
Result Code and Error Code (see RFC, [0]) for CDN.

However, regardless of sending 2,4 (``Insufficient resources to handle 
this operation now''), 2,6 (``A generic vendor-specific error occured in 
the LAC'', this is "Cisco style" from what we gathered using tcpdump), 
or 2,7 (``Try another''), it does not work: The Redbacks just ignore the 
other peers we're running, simply no more session is established, *even* 
on the machines *not* being in softshut.

So, e.g. we're running six peers, switching one of them into softshut 
seems the Redback make think *every* peer of ours is in softshut.

Did anybody discover similar problems? If yes, how did you solve this? 
Any clues what Redbacks ``want to hear'' to behave properly?

Best regards,

Bernd

--

[0] - http://tools.ietf.org/html/rfc2661

[1] - 
http://www.cisco.com/c/en/us/td/docs/ios/12_2sb/12_2sba/feature/guide/sba2tpef.pdf