24 Nov
2005
24 Nov
'05
12:12 p.m.
Hi Greger
Thanks for the reply. I will try the suggestions you have made. I am
using mhome=1. The trouble with the record_route_preset is that the
outgoing interface is assigned by dhcp, so I will have to try use dyndns I
think. I think it would be quite useful to be able to specify an interface
name instead of an address - have not tried this and don't know if it
could work. There are two RR entries in each forwarded message (one for
the internal interface and one for the external interface). I guess I
should do a trace and post that (with and without the if (uri!=myself) bit
around the loose_routing. I will try get that done today.
The solution I have seems to work ok, but as you say, I don't know what
problems I am making for myself futher along the line, so any feedback is
most welcome. Thanks for your interest.
The problem is that you are not loose routing messages that are destined for
your SER (uri!=myself). Loose routing means that the Route headers in
in-dialog messages are used for forwarding the message. By not loose
routing, but (probably) using lookup("location") or whatever you are doing
to the messages that should have been loose routed (ex. ACKs), you are
overriding the Routes. Your problem is that the routes are wrong in your
particular setup, not loose routing is hack that makes your SER behave
non-RFC compliant on Route headers and that may jump up and bite you down
the road...
g-)
Noel
Greger V. Teigre wrote:
Noel,
Thanks for taking the time to document your setup in such detail!
In general, the Vias and record-routes must be fixed. The rr parameter
enable_double_rr should be 1 (default). AFAIK, mhome=1 should handle the
vias. This thread may give some more background:
http://lists.iptel.org/pipermail/serdev/2003-April/000003.html
Have you tried using record_route_preset on outgoing (if I understand you
setup?) to add the correct (public) address?
Also, unless mhome is on, the source address of the incoming packet will
be used when sending out. This will be detected as a martian by the linux
kernel (see /var/log/messages). mhome will use the correct source address
and fix the vias.
Anyway, if your fix works, you should be fine, but remember that if you
change your setup later (or add a new GW, new UAs...), you may run into
problems as you are not following the RFC.
g-)
----- Original Message ----- From: "Noel Sharpe"
<noels(a)radnetwork.co.uk>
To: "Greger V. Teigre" <greger(a)teigre.com>
Cc: "'SER Users'" <serusers(a)lists.iptel.org>
Sent: Wednesday, November 16, 2005 9:52 AM
Subject: Re: [Serusers] Too Many Hops
Greger
I am trying to get a cascading hierarchy of geographically separated
SIP servers. The reason for this is that I am providing services to a
wireless community that connect to the internet via many different
gateways to the internet. Most of the users are behind a nat device,
and connect to a SER proxy on the wireless net work on the 10.0.0.0/8
network. Each SER server keeps a local location database, but forwards
any requests that it cannot complete to a sip proxy on the internet (to
which each SER proxy replicates all REGISTER's). It looks a bit like
this:
UAC(192.168.0.123)--->WIFI_AP(10.1.2.3)---->WIFI_GATEWAY_1--->INTERNET
for browsing / email / ftp etc
|
|
v
SER_SERVER on WiFi network
(10.0.0.123) also has (public IP Address)--->INTERNET GATEWAY for
VOIP 1--->SER_SERVER on internet
^
|
UAC(192.168.0.234)--->WIFI_AP(10.3.4.5)---->WIFI_GATEWAY_2--->INTERNET
for browsing / email / ftp etc |
| |
| |
v |
SER_SERVER on WiFi network
(10.3.4.123) |
also has (public
IP Address)--->INTERNET GATEWAY for VOIP 2-----
Each SER server implements it's own far end nat traversal based on
nathelper/rtpproxy, which means that expensive internet links are not
used to terminate calls between subscribers on the same wifi network.
The issue I was having with loose routing appeared to be because the
SER_SERVER on the WiFi network was multihomed. When a reply cam back
from the ser server on the internet (I think it was only ACK's causing
the problem), SER tried to loose_route the reply and the next hop was
infact the servers own internal ip address.
After loosing much hair, I worked out that if I included the condition I
could get the packet to route through to the correct final destination.
I am not sure if it's because there is a REAL issue with loose_route not
working with mhomed properly, or if it's a peculiarity of my setup.
Noel
Greger V. Teigre wrote:
Noel,
It would be interesting to get a description of the setup that caused
this problem. Was that Asterisk as well? Same setup as Giovanni?
Dependent on auth/non-auth and how Asterisk has been set up with SER,
we have seen different problems with signalling and proxying of RTP. I
know there was an lr vs. lr=on issue a while back, but somewhere we
don't have RFC compliant behavior (all messages with routes should be
loose routed).
g-)
----- Original Message ----- From: "Noel Sharpe"
<noels(a)radnetwork.co.uk>
To: "Giovanni Balasso" <giaso(a)yahoo.it>
Cc: <serusers(a)lists.iptel.org>
Sent: Tuesday, November 15, 2005 10:14 AM
Subject: Re: [Serusers] Too Many Hops
> sounds to me like you have a looping problem in your script. I had
> something similar when using the example from OnSIP.org. The
> loose_route bit needed to be inside a condition:
> if (uri!=myself){
> if (loose_route()) {
> route(1); }; };
> xlog/ngrep is your friend here as you will be able to see which
> message is being sent between the two servers.
>
> Noel
>
> Giovanni Balasso wrote:
>
>> Alle 09:47, mercoledì 02 novembre 2005, Matteo Piazza ha scritto:
>>
>>> I have Ser and asterisk on the same machine.
>>> When i try to call with a SIP phone registred on asterisk another
>>> sip
>>> phone also registred on asterisk through SER I receve this error
>>> message:
>>> Too many hops
>>>
>>
>> Too many hops is usually reached when there is no rule (or no way) to
>> deliver sip message, adding some log(), or better xlog(), to your
>> routing script could help you (and us) debugging and understanding
>> what's wrong, and which method(s) fail.
>>
>>
>>
>>> if (method == "INVITE") {
>>> if (uri =~"^sip:0[0-9]*@*"){
>>> log(1, "Check 1 succed Forwarding to Asterisk\n");
>>> rewritehostport("192.168.9.97:5061");
>>> t_relay();
>>> break;
>>> };
>>> };
>>>
>>>
>>
>> I don't think this will solve your problem but in my experience I had
>> better result with t_relay_to_udp("192.168.9.97","5061") than
>> rewritehostport("192.168.9.97:5061").
>>
>>
>
> _______________________________________________
> Serusers mailing list
> serusers(a)lists.iptel.org
> http://lists.iptel.org/mailman/listinfo/serusers
>