Scott Lawrence wrote: > On Fri, 2009-04-03 at 12:44 -0400, Damian Krzeminski wrote: >> Scott Lawrence wrote: >>> On Fri, 2009-04-03 at 10:48 -0400, Damian Krzeminski wrote: >>>> It might have been wrong but this is what it was. >>> It was wrong... not sure how we got away with it so long, but the issue >>> is correct. >>> >>> >> >> OK let's change it then. >> Just to make sure that we all understand how it's going to change: this is >> what's going to look like after I commit this patch >> >> Single server case (no change here): >> single server a.example.com (10.0.0.1) with registrar running on 5070 >> >> generated routeTo expression: >> >> <10.0.0.1:5070;transport=tcp;x-sipx-routetoreg> >> >> >> HA Case: >> >> a.example.com (10.0.0.1) with registrar running on 5070 >> b.example.com (10.0.0.2) with registrar running on 5070 >> >> >> a's routeTo: >> <rr.a.example.com;transport=tcp;x-sipx-routetoreg> >> >> b's routeTo: >> <rr.b.example.com;transport=tcp;x-sipx-routetoreg> >> >> >> The patch only changes SUBSCRIBE and default routing. Registrations are >> still sent to 10.0.0.1:5070 (for 'a') and 10.0.0.2:5070 (for 'b') >> >> >> Raymond: did I capture the changes correctly? >> Scott: makes sense? > > Registrations should always go to a name, never an IP address. >
I'll rewrite the patch to make it happen... > The current version of the sipx-dns script always generates the > rr.<fqdn> SRV record whether the system is HA or not, and it only > generates a TCP version, so all routing to the registry/redirect server > for any purpose can just be to: > > <rr.b.example.com;x-sipx-routetoreg> > > (with 'b.example.com' replaced with the fqdn of the current proxy) > > it's possible that systems that are being upgraded won't have that > record, in which case this won't work until it is installed, but I think > that the thing to do is to capture that in the release/upgrade notes. > > Alternatively, you could generate the registry routes with just the fqdn > and port 5070 for non-HA systems and the rr.<fqdn> SRV name for HA > systems. > > Let's do that - one system <fqdn>:<port>, more than one system rr.<fqdn> Can you clarify if we should generate 'transport=tcp' in a single system case? D. _______________________________________________ sipx-dev mailing list [email protected] List Archive: http://list.sipfoundry.org/archive/sipx-dev Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-dev
