Conference silence timeouts
Hi all. The solution to this one should be short. My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds). I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming. I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment. Anybody know how to increase the conference silence-hangup timeout? BB
Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps... -MC On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier <bjbrashier@gmail.com>wrote:
Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1). Any other ideas, or should I hunt through the code to try to override the behavior manually? BB On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org> wrote:
Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier <bjbrashier@gmail.com
wrote:
Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
On Thu, Aug 13, 2009 at 12:47 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
Get a packet capture and debug trace of the symptom occurring. Put the PCAP where we can download it and pastebin the debug log. We need to confirm who the culprit is and what events precipitate the disconnect. I'd also be curious to know if anyone else can reproduce these symptoms. BTW, which rev of FS are you running? Thanks, MC
I'm currently running current trunk synched up Tues morning, but it was happening in all of the versions I'd been using previous -- I first downloaded around the end of May. I'll look into getting you a PCap. I expected that this was a known thing with a parameter somewhere, so I haven't looked into it too terribly far myself, yet. I'm gonna try looking at the console outputs and logs myself, first. BB On Thu, Aug 13, 2009 at 12:44 PM, Michael Collins <msc@freeswitch.org>wrote:
On Thu, Aug 13, 2009 at 12:47 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
Get a packet capture and debug trace of the symptom occurring. Put the PCAP where we can download it and pastebin the debug log. We need to confirm who the culprit is and what events precipitate the disconnect. I'd also be curious to know if anyone else can reproduce these symptoms. BTW, which rev of FS are you running?
Thanks, MC
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
My guess is that its the other end killing the call due to rtp timeouts, not us killing the call. If you can confirm this the best method would be to get them not to do rtp timeouts. On Aug 13, 2009, at 1:47 PM, Bradley Brashier wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org> wrote: Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier <bjbrashier@gmail.com
wrote: Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_____
I had just thought of the exact same thing. I'm trying to test that now. Thanks for your input. BB On Thu, Aug 13, 2009 at 1:20 PM, Michael Jerris <mike@jerris.com> wrote:
My guess is that its the other end killing the call due to rtp timeouts, not us killing the call. If you can confirm this the best method would be to get them not to do rtp timeouts. On Aug 13, 2009, at 1:47 PM, Bradley Brashier wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org>wrote:
Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier < bjbrashier@gmail.com> wrote:
Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_____
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
I took a closer look at the SIP messages on the console. From it, I understand that it's not Freeswitch timing out, but rather FS is getting the "BYE" msg from somewhere else. I've tested phones and tested calling without going through the FS conference, though, and everything works fine. Then I saw something else odd inside the BYE msg: Reason: Q.850 ;cause=31 ;text="RTP Broken Connection" So I Googled "RTP Broken Connection" and saw several sites talking about AudioCodes gateways and Asterisk -- and our gateway is an AudioCodes. From these sites it sounds like AudioCodes is rather aggressive in detecting RTP breaks, and is interpreting the silence from FS as a break. So now I'm looking into ways to maybe send "I'm still here" RTP packets from FS or to tune the gateway to be less aggressive. I can't stop and get a clean packet capture right now because I've got a bunch of testers working on it today. I'll do that sometime when the system is less busy. BB On Thu, Aug 13, 2009 at 1:45 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I had just thought of the exact same thing. I'm trying to test that now. Thanks for your input.
BB
On Thu, Aug 13, 2009 at 1:20 PM, Michael Jerris <mike@jerris.com> wrote:
My guess is that its the other end killing the call due to rtp timeouts, not us killing the call. If you can confirm this the best method would be to get them not to do rtp timeouts. On Aug 13, 2009, at 1:47 PM, Bradley Brashier wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org>wrote:
Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier < bjbrashier@gmail.com> wrote:
Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_____
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
OK, I finally got a moment to do a packet capture and take a look at the streams. It became very clear very quickly that what happens is that during silence the gateway still sends RTP packets to Freeswitch, but Freeswitch doesn't send any back to the gateway. After 10s of this, the gateway says "Oh, the RPT must be broken" and it hangs up. We found a way to turn off this behavior in the gateway, and the good news is that it did indeed fix the problem. But we'd rather not rely on that as a long-term solution because then we can't detect and drop RTP streams that really are broken. So now I'm back to looking at Freeswitch to figure out how to send just a single packet every second or so during silence. If anyone knows of a way to do this, let me know, otherwise I'll get back to you if and when I find one. BB On Thu, Aug 13, 2009 at 2:48 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I took a closer look at the SIP messages on the console. From it, I understand that it's not Freeswitch timing out, but rather FS is getting the "BYE" msg from somewhere else. I've tested phones and tested calling without going through the FS conference, though, and everything works fine. Then I saw something else odd inside the BYE msg:
Reason: Q.850 ;cause=31 ;text="RTP Broken Connection" So I Googled "RTP Broken Connection" and saw several sites talking about AudioCodes gateways and Asterisk -- and our gateway is an AudioCodes. From these sites it sounds like AudioCodes is rather aggressive in detecting RTP breaks, and is interpreting the silence from FS as a break.
So now I'm looking into ways to maybe send "I'm still here" RTP packets from FS or to tune the gateway to be less aggressive. I can't stop and get a clean packet capture right now because I've got a bunch of testers working on it today. I'll do that sometime when the system is less busy.
BB
On Thu, Aug 13, 2009 at 1:45 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I had just thought of the exact same thing. I'm trying to test that now. Thanks for your input.
BB
On Thu, Aug 13, 2009 at 1:20 PM, Michael Jerris <mike@jerris.com>wrote:
My guess is that its the other end killing the call due to rtp timeouts, not us killing the call. If you can confirm this the best method would be to get them not to do rtp timeouts. On Aug 13, 2009, at 1:47 PM, Bradley Brashier wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org>wrote:
Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier < bjbrashier@gmail.com> wrote:
Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_____
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
My suggestion is to use sip session timers not rtp timeouts as rtp is supposed to be discontinuous. That being said, we have several settings to continuously send media, but then you are doing exactly what you said you didn't want to do. Mike On Aug 13, 2009, at 6:24 PM, Bradley Brashier wrote:
OK, I finally got a moment to do a packet capture and take a look at the streams. It became very clear very quickly that what happens is that during silence the gateway still sends RTP packets to Freeswitch, but Freeswitch doesn't send any back to the gateway. After 10s of this, the gateway says "Oh, the RPT must be broken" and it hangs up.
We found a way to turn off this behavior in the gateway, and the good news is that it did indeed fix the problem. But we'd rather not rely on that as a long-term solution because then we can't detect and drop RTP streams that really are broken.
So now I'm back to looking at Freeswitch to figure out how to send just a single packet every second or so during silence. If anyone knows of a way to do this, let me know, otherwise I'll get back to you if and when I find one.
BB
On Thu, Aug 13, 2009 at 2:48 PM, Bradley Brashier <bjbrashier@gmail.com
wrote: I took a closer look at the SIP messages on the console. From it, I understand that it's not Freeswitch timing out, but rather FS is getting the "BYE" msg from somewhere else. I've tested phones and tested calling without going through the FS conference, though, and everything works fine. Then I saw something else odd inside the BYE msg:
Reason: Q.850 ;cause=31 ;text="RTP Broken Connection" So I Googled "RTP Broken Connection" and saw several sites talking about AudioCodes gateways and Asterisk -- and our gateway is an AudioCodes. From these sites it sounds like AudioCodes is rather aggressive in detecting RTP breaks, and is interpreting the silence from FS as a break.
So now I'm looking into ways to maybe send "I'm still here" RTP packets from FS or to tune the gateway to be less aggressive. I can't stop and get a clean packet capture right now because I've got a bunch of testers working on it today. I'll do that sometime when the system is less busy.
BB
On Thu, Aug 13, 2009 at 1:45 PM, Bradley Brashier <bjbrashier@gmail.com
wrote: I had just thought of the exact same thing. I'm trying to test that now. Thanks for your input.
BB
On Thu, Aug 13, 2009 at 1:20 PM, Michael Jerris <mike@jerris.com> wrote: My guess is that its the other end killing the call due to rtp timeouts, not us killing the call. If you can confirm this the best method would be to get them not to do rtp timeouts.
On Aug 13, 2009, at 1:47 PM, Bradley Brashier wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org> wrote: Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier <bjbrashier@gmail.com
wrote: Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_____
I didn't see any SIP session timers in the wiki. Since I'm already using the event socket for control, my current plan is to use sched_api to play a file with a short (20ms?) clip of silence, capture the play_file event and use it to reschule another one for a couple of seconds later. I'll let you know what happens. BB On Thu, Aug 13, 2009 at 10:47 PM, Michael Jerris <mike@jerris.com> wrote:
My suggestion is to use sip session timers not rtp timeouts as rtp is supposed to be discontinuous. That being said, we have several settings to continuously send media, but then you are doing exactly what you said you didn't want to do. Mike
On Aug 13, 2009, at 6:24 PM, Bradley Brashier wrote:
OK, I finally got a moment to do a packet capture and take a look at the streams. It became very clear very quickly that what happens is that during silence the gateway still sends RTP packets to Freeswitch, but Freeswitch doesn't send any back to the gateway. After 10s of this, the gateway says "Oh, the RPT must be broken" and it hangs up.
We found a way to turn off this behavior in the gateway, and the good news is that it did indeed fix the problem. But we'd rather not rely on that as a long-term solution because then we can't detect and drop RTP streams that really are broken.
So now I'm back to looking at Freeswitch to figure out how to send just a single packet every second or so during silence. If anyone knows of a way to do this, let me know, otherwise I'll get back to you if and when I find one.
BB
On Thu, Aug 13, 2009 at 2:48 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I took a closer look at the SIP messages on the console. From it, I understand that it's not Freeswitch timing out, but rather FS is getting the "BYE" msg from somewhere else. I've tested phones and tested calling without going through the FS conference, though, and everything works fine. Then I saw something else odd inside the BYE msg:
Reason: Q.850 ;cause=31 ;text="RTP Broken Connection" So I Googled "RTP Broken Connection" and saw several sites talking about AudioCodes gateways and Asterisk -- and our gateway is an AudioCodes. From these sites it sounds like AudioCodes is rather aggressive in detecting RTP breaks, and is interpreting the silence from FS as a break.
So now I'm looking into ways to maybe send "I'm still here" RTP packets from FS or to tune the gateway to be less aggressive. I can't stop and get a clean packet capture right now because I've got a bunch of testers working on it today. I'll do that sometime when the system is less busy.
BB
On Thu, Aug 13, 2009 at 1:45 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I had just thought of the exact same thing. I'm trying to test that now. Thanks for your input.
BB
On Thu, Aug 13, 2009 at 1:20 PM, Michael Jerris <mike@jerris.com>wrote:
My guess is that its the other end killing the call due to rtp timeouts, not us killing the call. If you can confirm this the best method would be to get them not to do rtp timeouts. On Aug 13, 2009, at 1:47 PM, Bradley Brashier wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org>wrote:
Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier < bjbrashier@gmail.com> wrote:
Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_____
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
That sounds horrible. There are settings both in sip/rtp and in conference to do this already. http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#enable-timer http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#session-timeout http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#rtp-timeout-sec http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#rtp-hold-timeout-s... http://wiki.freeswitch.org/wiki/VAD_and_CNG Mike On Aug 14, 2009, at 2:41 PM, Bradley Brashier wrote:
I didn't see any SIP session timers in the wiki. Since I'm already using the event socket for control, my current plan is to use sched_api to play a file with a short (20ms?) clip of silence, capture the play_file event and use it to reschule another one for a couple of seconds later.
I'll let you know what happens.
BB
On Thu, Aug 13, 2009 at 10:47 PM, Michael Jerris <mike@jerris.com> wrote: My suggestion is to use sip session timers not rtp timeouts as rtp is supposed to be discontinuous. That being said, we have several settings to continuously send media, but then you are doing exactly what you said you didn't want to do.
Mike
On Aug 13, 2009, at 6:24 PM, Bradley Brashier wrote:
OK, I finally got a moment to do a packet capture and take a look at the streams. It became very clear very quickly that what happens is that during silence the gateway still sends RTP packets to Freeswitch, but Freeswitch doesn't send any back to the gateway. After 10s of this, the gateway says "Oh, the RPT must be broken" and it hangs up.
We found a way to turn off this behavior in the gateway, and the good news is that it did indeed fix the problem. But we'd rather not rely on that as a long-term solution because then we can't detect and drop RTP streams that really are broken.
So now I'm back to looking at Freeswitch to figure out how to send just a single packet every second or so during silence. If anyone knows of a way to do this, let me know, otherwise I'll get back to you if and when I find one.
BB
On Thu, Aug 13, 2009 at 2:48 PM, Bradley Brashier <bjbrashier@gmail.com
wrote: I took a closer look at the SIP messages on the console. From it, I understand that it's not Freeswitch timing out, but rather FS is getting the "BYE" msg from somewhere else. I've tested phones and tested calling without going through the FS conference, though, and everything works fine. Then I saw something else odd inside the BYE msg:
Reason: Q.850 ;cause=31 ;text="RTP Broken Connection" So I Googled "RTP Broken Connection" and saw several sites talking about AudioCodes gateways and Asterisk -- and our gateway is an AudioCodes. From these sites it sounds like AudioCodes is rather aggressive in detecting RTP breaks, and is interpreting the silence from FS as a break.
So now I'm looking into ways to maybe send "I'm still here" RTP packets from FS or to tune the gateway to be less aggressive. I can't stop and get a clean packet capture right now because I've got a bunch of testers working on it today. I'll do that sometime when the system is less busy.
BB
On Thu, Aug 13, 2009 at 1:45 PM, Bradley Brashier <bjbrashier@gmail.com
wrote: I had just thought of the exact same thing. I'm trying to test that now. Thanks for your input.
BB
On Thu, Aug 13, 2009 at 1:20 PM, Michael Jerris <mike@jerris.com> wrote: My guess is that its the other end killing the call due to rtp timeouts, not us killing the call. If you can confirm this the best method would be to get them not to do rtp timeouts.
On Aug 13, 2009, at 1:47 PM, Bradley Brashier wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org> wrote: Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier <bjbrashier@gmail.com
wrote: Hi all.
The solution to this one should be short.
My conference hangs up when there's 2+ users and silence for 5 sec or so. I'm trying to find a parameter that changes that (I'd rather it be, say, 60 seconds).
I didn't see a parameter like this specific to conferences, so I looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set to 300 (the default), so I'm pretty sure that's not the problem. I also searched through the mod_conference.c code and didn't see it, though I was only skimming.
I'm not 100% convinced that this is limited to conferences, but I don't currently have a way to test in a non-conference environment.
Anybody know how to increase the conference silence-hangup timeout?
BB
_____
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
All right, I'm confused. The RTP timeout parameters have no documentation, and from the names, I'd have guessed that they hang up after a specified amount of time, not send some other signal. The session-timeout timer talks about calls expiring, and sending another SIP invite, which I don't think is what I need. It also says it shouldn't be set to less than 30 minutes. I've been poking through RFC 4028, which the enable-timer parameter mentions, but I'm not sure those will be helpful, either, because I've got the SIP and RTP traffic going on different paths, and it's the RTP that's having problems. Still, maybe I can use one of the timers to do something different, but then I'm not sure how that's any better than using the event socket the way I planned. Clearly I'm missing something. BB On Fri, Aug 14, 2009 at 12:17 PM, Michael Jerris <mike@jerris.com> wrote:
That sounds horrible. There are settings both in sip/rtp and in conference to do this already. http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#enable-timer http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#session-timeout http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#rtp-timeout-sec
http://wiki.freeswitch.org/wiki/Sofia_Configuration_Files#rtp-hold-timeout-s... http://wiki.freeswitch.org/wiki/VAD_and_CNG
Mike
On Aug 14, 2009, at 2:41 PM, Bradley Brashier wrote:
I didn't see any SIP session timers in the wiki. Since I'm already using the event socket for control, my current plan is to use sched_api to play a file with a short (20ms?) clip of silence, capture the play_file event and use it to reschule another one for a couple of seconds later.
I'll let you know what happens.
BB
On Thu, Aug 13, 2009 at 10:47 PM, Michael Jerris <mike@jerris.com> wrote:
My suggestion is to use sip session timers not rtp timeouts as rtp is supposed to be discontinuous. That being said, we have several settings to continuously send media, but then you are doing exactly what you said you didn't want to do. Mike
On Aug 13, 2009, at 6:24 PM, Bradley Brashier wrote:
OK, I finally got a moment to do a packet capture and take a look at the streams. It became very clear very quickly that what happens is that during silence the gateway still sends RTP packets to Freeswitch, but Freeswitch doesn't send any back to the gateway. After 10s of this, the gateway says "Oh, the RPT must be broken" and it hangs up.
We found a way to turn off this behavior in the gateway, and the good news is that it did indeed fix the problem. But we'd rather not rely on that as a long-term solution because then we can't detect and drop RTP streams that really are broken.
So now I'm back to looking at Freeswitch to figure out how to send just a single packet every second or so during silence. If anyone knows of a way to do this, let me know, otherwise I'll get back to you if and when I find one.
BB
On Thu, Aug 13, 2009 at 2:48 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I took a closer look at the SIP messages on the console. From it, I understand that it's not Freeswitch timing out, but rather FS is getting the "BYE" msg from somewhere else. I've tested phones and tested calling without going through the FS conference, though, and everything works fine. Then I saw something else odd inside the BYE msg:
Reason: Q.850 ;cause=31 ;text="RTP Broken Connection" So I Googled "RTP Broken Connection" and saw several sites talking about AudioCodes gateways and Asterisk -- and our gateway is an AudioCodes. From these sites it sounds like AudioCodes is rather aggressive in detecting RTP breaks, and is interpreting the silence from FS as a break.
So now I'm looking into ways to maybe send "I'm still here" RTP packets from FS or to tune the gateway to be less aggressive. I can't stop and get a clean packet capture right now because I've got a bunch of testers working on it today. I'll do that sometime when the system is less busy.
BB
On Thu, Aug 13, 2009 at 1:45 PM, Bradley Brashier <bjbrashier@gmail.com>wrote:
I had just thought of the exact same thing. I'm trying to test that now. Thanks for your input.
BB
On Thu, Aug 13, 2009 at 1:20 PM, Michael Jerris <mike@jerris.com>wrote:
My guess is that its the other end killing the call due to rtp timeouts, not us killing the call. If you can confirm this the best method would be to get them not to do rtp timeouts. On Aug 13, 2009, at 1:47 PM, Bradley Brashier wrote:
I'm sure that would work, but I'm worried about it sucking up bandwidth, especially since you'd need it on every caller (since otherwise the one person who had it could hang up and you'd be back to square 1).
Any other ideas, or should I hunt through the code to try to override the behavior manually?
BB
On Thu, Aug 13, 2009 at 9:50 AM, Michael Collins <msc@freeswitch.org>wrote:
Check out the 'waste' member flag. I think if at least one member has that set then RTP will get sent out even during silence. Let us know if that helps...
-MC
On Thu, Aug 13, 2009 at 11:37 AM, Bradley Brashier < bjbrashier@gmail.com> wrote:
> Hi all. > > The solution to this one should be short. > > My conference hangs up when there's 2+ users and silence for 5 sec or > so. I'm trying to find a parameter that changes that (I'd rather it be, > say, 60 seconds). > > I didn't see a parameter like this specific to conferences, so I > looked abroad a bit. I found rtp-timeout-sec in sip_profiles, but it's set > to 300 (the default), so I'm pretty sure that's not the problem. I also > searched through the mod_conference.c code and didn't see it, though I was > only skimming. > > I'm not 100% convinced that this is limited to conferences, but I > don't currently have a way to test in a non-conference environment. > > Anybody know how to increase the conference silence-hangup timeout? > > BB > > _____
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
_______________________________________________ FreeSWITCH-users mailing list FreeSWITCH-users@lists.freeswitch.org http://lists.freeswitch.org/mailman/listinfo/freeswitch-users UNSUBSCRIBE:http://lists.freeswitch.org/mailman/options/freeswitch-users http://www.freeswitch.org
participants (3)
-
Bradley Brashier
-
Michael Collins
-
Michael Jerris