Re: [sipcore] Bypassing out-of-service intermediate Proxy

[<[e-mail]>]

I understand that the usage of DNS to provide alternate server is a way to ensure redundancy. In such solution, to be able to handle SIP requests inside a SIP dialog and for which the proxy need to be "dialog statefull", there is a need to implement a mechanism that would allow the "alternate" server to share the same SIP dialog contexts.

My idea is in case of unavailability of the next hop and in case no alternate server is returned by DNS (or alternate servers are unreachable too), instead of rejecting the request (which would make the call/session fails), the SIP Proxy may send it to the hop after the unreachable one in the Route. This alternative would be authorized only if the Proxy knows that it's better to bypass the unreachable node instead of rejecting the request. The SIP proxy may get this information via a URI parameter in the Route entry of the "unreachable" node for example

Best regards,

Youssef

-----Message d'origine-----
De : Worley, Dale R (Dale) [mailto:[e-mail]] 
Envoyé : vendredi 29 octobre 2010 20:22
À : CHADLI Youssef RD-CORE-ISS; [e-mail]
Cc : [e-mail]
Objet : RE: [sipcore] Bypassing out-of-service intermediate Proxy

________________________________________
From: [e-mail] [[e-mail]] On Behalf Of [e-mail] [[e-mail]]

I understand that the intention of this text in RFC 3261 is to fail over to an alternate server that serves the same function. Though, it's not explicitly stated.

However, I did not find any text in RFC 3261 that would make forbidden failing over to a proxy further down the path.  Moreover, I did not see any problem to such behaviour.
________________________________

If a particular URI is included in a Route header, then there is some important function the server with that URI is to perform.  If one element ignores this URI and sends the request to the next URI in the Route header, presumably the function of the skipped server will not be performed.

In all cases that I have heard about, the correct way to provide alternate servers for a failed server is to construct a DNS SRV name which maps to the primary and alternate server as is desired.

For instance, if functionA is to be serviced by serverA but in an emergency the request can be routed to serverB, and if functionB is serviced by serverB, then we construct two SRV names:

functionA    SRV   serverA  priority=1
functionA    SRV   serverB  priority=2

functionB    SRV   serverB priority=1

and use this Route header:

Route:  sip:functionA, sip:functionB

If an element has a request with this Route header, it will send the request to sp:functionA using the RFC 3263 rules, viz., attempt to send to serverA, and if that fails, attempt to send to serverB.

Dale