Hi,
I have a simple kamailio install (2 servers, using location service and a failover node with dispatcher, STUN, clients behind different NATs), without rtpproxy, only peer-to-peer RTP and TURN server if the connection is really messy (it's not relevant here). Signaling is over TLS. Both of the clients are behind NAT.
Basically everything works, but some of the calls are dropped after 15 minutes and some seconds, for me it seems the RTP connection is dropped but not sure.
I cannot find find out wether it's a kamailio problem or client/NAT problem.
Has anyone idea what is going on?
Thanks, Andras
Hello,
I expect that the signaling is ok at least for call setup.
From signling point of view, I can think of following situations: - endpoints send keep alive packets (or session updates) which are no answered. You can add an xlog(...) at the top of request_route{} and reply_route{} blocks printing at least the method, call-id, cseq, from and to header, plus the response code for reply block. In this case you can see if there is some signaling before call is dropped.
- the tls connection is cut (can be done by routers in the middle) and one endpoint disconnects the call instead of trying first to reconnect
If none of the above is the case, then might be some problem with rtp transmission and call is disconnected by a device because of that.
Cheers, Daniel
On 16/07/14 20:39, Andras FOGARASI wrote:
Hi,
I have a simple kamailio install (2 servers, using location service and a failover node with dispatcher, STUN, clients behind different NATs), without rtpproxy, only peer-to-peer RTP and TURN server if the connection is really messy (it's not relevant here). Signaling is over TLS. Both of the clients are behind NAT.
Basically everything works, but some of the calls are dropped after 15 minutes and some seconds, for me it seems the RTP connection is dropped but not sure.
I cannot find find out wether it's a kamailio problem or client/NAT problem.
Has anyone idea what is going on?
Thanks, Andras
SIP Express Router (SER) and Kamailio (OpenSER) - sr-users mailing list sr-users@lists.sip-router.org http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-users
On Jul 16, 2014, at 3:54 PM, Daniel-Constantin Mierla miconda@gmail.com wrote:
Hello,
I expect that the signaling is ok at least for call setup.
From signling point of view, I can think of following situations:
- endpoints send keep alive packets (or session updates) which are no answered. You can add an xlog(...) at the top of request_route{} and reply_route{} blocks printing at least the method, call-id, cseq, from and to header, plus the response code for reply block. In this case you can see if there is some signaling before call is dropped.
Is this happening just on calls between two phones in your domain, or is there a carrier/federation involved?
--FC
On 7/16/14, 10:00 PM, Frank Carmickle wrote:
On Jul 16, 2014, at 3:54 PM, Daniel-Constantin Mierla miconda@gmail.com wrote:
Hello,
I expect that the signaling is ok at least for call setup.
From signling point of view, I can think of following situations:
- endpoints send keep alive packets (or session updates) which are no answered. You can add an xlog(...) at the top of request_route{} and reply_route{} blocks printing at least the method, call-id, cseq, from and to header, plus the response code for reply block. In this case you can see if there is some signaling before call is dropped.
Is this happening just on calls between two phones in your domain, or is there a carrier/federation involved?
No other parties are involved, only the two phones involved (and the proxy of course).
Andras
On Jul 16, 2014, at 4:05 PM, Andras FOGARASI fogarasi@fogarasi.com wrote:
On 7/16/14, 10:00 PM, Frank Carmickle wrote:
On Jul 16, 2014, at 3:54 PM, Daniel-Constantin Mierla miconda@gmail.com wrote:
Hello,
I expect that the signaling is ok at least for call setup.
From signling point of view, I can think of following situations:
- endpoints send keep alive packets (or session updates) which are no answered. You can add an xlog(...) at the top of request_route{} and reply_route{} blocks printing at least the method, call-id, cseq, from and to header, plus the response code for reply block. In this case you can see if there is some signaling before call is dropped.
Is this happening just on calls between two phones in your domain, or is there a carrier/federation involved?
No other parties are involved, only the two phones involved (and the proxy of course).
I would expect that if it was a NAT issue you would see it much sooner than 15 minutes, 30-60 seconds. Are session timers being stripped by Kamailio? You say it's a TURN server or is it acting more like a media relay where it is signaled into the path? What TURN server are you using? How is it configured?
--FC
On 7/17/14, 3:41 PM, Frank Carmickle wrote:
On Jul 16, 2014, at 4:05 PM, Andras FOGARASI fogarasi@fogarasi.com wrote:
On 7/16/14, 10:00 PM, Frank Carmickle wrote:
On Jul 16, 2014, at 3:54 PM, Daniel-Constantin Mierla miconda@gmail.com wrote:
Hello,
I expect that the signaling is ok at least for call setup.
From signling point of view, I can think of following situations:
- endpoints send keep alive packets (or session updates) which are no answered. You can add an xlog(...) at the top of request_route{} and reply_route{} blocks printing at least the method, call-id, cseq, from and to header, plus the response code for reply block. In this case you can see if there is some signaling before call is dropped.
Is this happening just on calls between two phones in your domain, or is there a carrier/federation involved?
No other parties are involved, only the two phones involved (and the proxy of course).
I would expect that if it was a NAT issue you would see it much sooner than 15 minutes, 30-60 seconds. Are session timers being stripped by Kamailio? You say it's a TURN server or is it acting more like a media relay where it is signaled into the path? What TURN server are you using? How is it configured?
The problem occurs even without TURN, in pure peer-to-peer mode. We use TURN only in emergency case (symmetric NAT and like that...). I do nothing with session timers - i didn't think about it until now...
Andras
On Jul 17, 2014, at 11:27 AM, Andras FOGARASI fogarasi@fogarasi.com wrote:
On 7/17/14, 3:41 PM, Frank Carmickle wrote:
I would expect that if it was a NAT issue you would see it much sooner than 15 minutes, 30-60 seconds. Are session timers being stripped by Kamailio? You say it's a TURN server or is it acting more like a media relay where it is signaled into the path? What TURN server are you using? How is it configured?
The problem occurs even without TURN, in pure peer-to-peer mode. We use TURN only in emergency case (symmetric NAT and like that...). I do nothing with session timers - i didn't think about it until now…
How do you set up the TURN only in emergency case? Do the phones do it themselves? Does Kamailio control the TURN, rtpengine/mediaproxy-ng?
Who sends the bye?
--FC
On 7/17/14, 5:34 PM, Frank Carmickle wrote:
On Jul 17, 2014, at 11:27 AM, Andras FOGARASI fogarasi@fogarasi.com wrote:
On 7/17/14, 3:41 PM, Frank Carmickle wrote:
I would expect that if it was a NAT issue you would see it much sooner than 15 minutes, 30-60 seconds. Are session timers being stripped by Kamailio? You say it's a TURN server or is it acting more like a media relay where it is signaled into the path? What TURN server are you using? How is it configured?
The problem occurs even without TURN, in pure peer-to-peer mode. We use TURN only in emergency case (symmetric NAT and like that...). I do nothing with session timers - i didn't think about it until now…
How do you set up the TURN only in emergency case? Do the phones do it themselves? Does Kamailio control the TURN, rtpengine/mediaproxy-ng?
Who sends the bye?
Caller send BYE as i see.
Meanwhile i made some debug: after 15 minutes an UPDATE comes. Sometimes the UPDATE is answered with 200 OK, then the call doesn't drop, continues and everything works as expected. Sometimes it's not answered at all (it seems timeout), then hangs up.
Andras
On 17/07/14 20:56, Andras FOGARASI wrote:
On 7/17/14, 5:34 PM, Frank Carmickle wrote:
On Jul 17, 2014, at 11:27 AM, Andras FOGARASI fogarasi@fogarasi.com wrote:
On 7/17/14, 3:41 PM, Frank Carmickle wrote:
I would expect that if it was a NAT issue you would see it much sooner than 15 minutes, 30-60 seconds. Are session timers being stripped by Kamailio? You say it's a TURN server or is it acting more like a media relay where it is signaled into the path? What TURN server are you using? How is it configured?
The problem occurs even without TURN, in pure peer-to-peer mode. We use TURN only in emergency case (symmetric NAT and like that...). I do nothing with session timers - i didn't think about it until now…
How do you set up the TURN only in emergency case? Do the phones do it themselves? Does Kamailio control the TURN, rtpengine/mediaproxy-ng?
Who sends the bye?
Caller send BYE as i see.
Meanwhile i made some debug: after 15 minutes an UPDATE comes. Sometimes the UPDATE is answered with 200 OK, then the call doesn't drop, continues and everything works as expected. Sometimes it's not answered at all (it seems timeout), then hangs up.
Is the UPDATE forwarded from the proxy? Can you see some activity on the network? Maybe you can run kamailio with debug=3 in cfg and get more log messages from there that can provide further hints.
Also, you can try to run over tcp to see if it reproduces in the same way.
Cheers, Daniel
-
Andras FOGARASI -
Daniel-Constantin Mierla -
Frank Carmickle