[OpenSIPS-Users] Is it a kind of TCP keep alive produced by OpenSIPS?

Thu Oct 27 09:47:34 CEST 2016

Hi, Rodrigo!

See my answers inline.

BR

Răzvan Crainea
OpenSIPS Solutions
www.opensips-solutions.com

On 10/26/2016 08:15 PM, Rodrigo Pimenta Carvalho wrote:
>
> Hi Răzvan.
>
>
> Thank you very much.
>
> I'm facing a problem here related to TCP connection teared down during 
> dialogs.
>
> While a peer is not in dialogs, its TCP connection to OpenSIPS keeps 
> online all the time.
>
> However, when such peer enters in a conversation (be part of a 
> dialog), after few minutes there is a EOF received in a socket. After 
> this, OpenSIPS can no more send SIP BYEs to the respective peer. In 
> the log I can see:
>
>
> Jan 02 01:38:45 colibri-imx6-jfl opensips[21018]: Jan  2 01:38:45 
> [21027] DBG:core:tcp_read: EOF on 0x74e3d048, FD 24
> Jan 02 01:38:45 colibri-imx6-jfl opensips[21018]: Jan  2 01:38:45 
> [21027] DBG:core:tcp_read_req: EOF received
> Jan 02 01:38:45 colibri-imx6-jfl opensips[21018]: Jan  2 01:38:45 
> [21027] DBG:core:io_watch_del: [TCP_worker] io_watch_del op on index 0 
> 24 (0x1875e8, 24, 0, 0x10,0x3) fd_no=3 called
> Jan 02 01:38:45 colibri-imx6-jfl opensips[21018]: Jan  2 01:38:45 
> [21027] DBG:core:tcpconn_release:  releasing con 0x74e3d048, state -1, 
> fd=-1, id=3
> Jan 02 01:38:45 colibri-imx6-jfl opensips[21018]: Jan  2 01:38:45 
> [21027] DBG:core:tcpconn_release:  extra_data (nil)
> Jan 02 01:38:45 colibri-imx6-jfl opensips[21018]: Jan  2 01:38:45 
> [21029] DBG:core:handle_tcp_worker: reader response= 74e3d048, -1 from 2
> Jan 02 01:38:45 colibri-imx6-jfl opensips[21018]: Jan  2 01:38:45 
> [21029] DBG:core:tcpconn_destroy: destroying connection 0x74e3d048, 
> flags 0006
>
> ...
>
> When OpenSIPS try to send a SIP BYE via socket 0x74e3d048 , I can see 
> the log:
>
> Jan 02 01:40:49 colibri-imx6-jfl opensips[21018]: Jan  2 01:40:49 
> [21026] DBG:core:proto_tcp_send: no open tcp connection found, opening 
> new one, async = 1
>
>
> I have already used the flag "Pp" in the creation of dialogs, but it 
> didn't take effect. That is, even with "Pp" I'm still getting "EOF" in 
> the TCP socket.
>
>
> 1 - Should the flag "Pp" avoid those EOFs during dialogs?
>
Ideally, it should. However if the client does not "like" the pinging 
and closes the connection, there's not that much we can do.
>
>
> That flag causes the OpenSIPS to send SIP OPTIONS. The peers are 
> replying with SIP 500.
>
That's not really an issue. The SIP OPTIONs pinging has two purposes: 1. 
verify if the dialog is still active, and 2. keep the NAT pinhole open. 
If the SIP client doesn't know how to reply to in-dialog pinging, then 
1. isn't really useful. So the reply code doesn't really matter, unless 
it is a 408, which means that the peer did not respond at all, and the 
dialog will be turn down.
>
>
> 2- Is a SIP 500 reply enough to OpenSIPS keep the dialog connected?
>
Any communication between OpenSIPS and the client keeps the NAT pinhole 
open (see 2. above). From SIP perspective, that 500 could have a lot of 
meanings: the client does not know how to reply, or there was an 
internal error that could not process the message. However, this whole 
communication will keep the connection alive.
>
>
> 3 - Does it make sense getting absence of keep alive messages during 
> dialogs?
>
So as I said above, any pinging method is useful to keep the connection 
alive. Regarding the dialogs, there is no special need for pinging from 
SIP perspective, unless you are using SST, which I presume you are not.
>
>
> Any hint will be very helpful!
>
> P.S.: I will check the TCP trace too, looking for keep alives.
>
 From your diagnosis, I understand that the client closes the 
connection, right? Unfortunately I don't know why it does that, perhaps 
the best thing to do is to check the client's logs.
>
>
> Best regards.
>
>
>
> RODRIGO PIMENTA CARVALHO
> Inatel Competence Center
> Software
> Ph: +55 35 3471 9200 RAMAL 979
>
>
> ------------------------------------------------------------------------
> *De:* users-bounces at lists.opensips.org 
> <users-bounces at lists.opensips.org> em nome de Răzvan Crainea 
> <razvan at opensips.org>
> *Enviado:* quarta-feira, 26 de outubro de 2016 13:08
> *Para:* users at lists.opensips.org
> *Assunto:* Re: [OpenSIPS-Users] Is it a kind of TCP keep alive 
> produced by OpenSIPS?
> Hi, Rodrigo!
>
> The logs you are tracing are printed when OpenSIPS receives something 
> from the client, and then immediately responds back. Due to the fact 
> that we don't see any other debug messages, like SIP parsing & stuff, 
> makes me think that it is a CRLF pinging - the client periodically 
> sends a CRLFCRLF TCP message to OpenSIPS, and OpenSIPS responds with a 
> single CRLF. Note that this is different from a TCP keep-alive, where 
> each peer send a 0-length TCP message, without any body. That message 
> doesn't even get to the application layer.
> However, tracing the communication between OpenSIPS and the client 
> should confirm the above :).
>
> Best regards,
> Răzvan Crainea
> OpenSIPS Solutions
> www.opensips-solutions.com
> Home — OpenSIPS Solutions <http://www.opensips-solutions.com/>
> www.opensips-solutions.com
> OpenSIPS is a mature Open Source implementation of a SIP server. 
> OpenSIPS is more than a SIP proxy/router as it includes 
> application-level functionalities.
>
> On 10/26/2016 05:10 PM, Rodrigo Pimenta Carvalho wrote:
>>
>> Dear OpenSIPS users,
>>
>>
>> In the OpenSIPS log I see:
>>
>>
>> Jan 01 19:30:38 colibri-imx6-jfl opensips[3444]: Jan  1 19:30:38 
>> [3451] DBG:core:tcp_read_req: Using the global ( per process ) buff
>> Jan 01 19:30:38 colibri-imx6-jfl opensips[3444]: Jan 1 19:30:38 
>> [3451] DBG:core:tcp_handle_req: content-length= 0
>> Jan 01 19:30:38 colibri-imx6-jfl opensips[3444]: Jan 1 19:30:38 
>> [3451] DBG:core:async_tsend_stream: Async successful write from first 
>> try on 0x74e13548
>> Jan 01 19:30:38 colibri-imx6-jfl opensips[3444]: Jan 1 19:30:38 
>> [3451] DBG:core:tcp_read_req: tcp_read_req end
>>
>> The frequency is 1 time at each 1,5 minute. There is only one client 
>> online. I suspect that OpenSIPS uses the socket 0x74e13548 to send 
>> messages to such client. The client became online using TCP.
>>
>> Just to confirm, is this log a result of a TCP keep alive function 
>> enabled?
>>
>> Best regards.
>>
>>
>>
>>
>> RODRIGO PIMENTA CARVALHO
>> Inatel Competence Center
>> Software
>> Ph: +55 35 3471 9200 RAMAL 979
>>
>>
>> _______________________________________________
>> Users mailing list
>> Users at lists.opensips.org
>> http://lists.opensips.org/cgi-bin/mailman/listinfo/users
>
>
>
> _______________________________________________
> Users mailing list
> Users at lists.opensips.org
> http://lists.opensips.org/cgi-bin/mailman/listinfo/users

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensips.org/pipermail/users/attachments/20161027/8ee5a2d0/attachment-0001.htm>