[OpenSIPS-Users] Registration Loss on opensips SBC

Bogdan-Andrei Iancu bogdan at opensips.org
Tue Feb 6 07:12:31 EST 2018


Hi Royee,

So, my primary suspicion on the save() points to the event routes - to 
confirm this, you can use the benchmark module to log the required time 
to execute those routes. And whenever you have an alert on save(), you 
can correlate it with the execution time for the event routes.

On the t_relay() part, it may be more complicated, as the TM module has 
a complex system of callbacks that are triggered (inline) . And there 
are many modules using the TM callbacks , any of them potentially 
delaying the t_relay() action. Another reason may be DNS lookups, if 
your relaying is DNS based (and not IP based).

Regards,

Bogdan-Andrei Iancu

OpenSIPS Founder and Developer
   http://www.opensips-solutions.com
OpenSIPS Summit 2018
   http://www.opensips.org/events/Summit-2018Amsterdam

On 02/06/2018 10:30 AM, Royee Tichauer wrote:
> Thanks for lead Bogdan we will look into it.
>
> Do you also have any idea why the t_relay was blocking for so long?
>
> Royee
>
> On Mon, Feb 5, 2018 at 7:23 PM Bogdan-Andrei Iancu 
> <bogdan at opensips.org <mailto:bogdan at opensips.org>> wrote:
>
>     Hi Royee,
>
>     I don't think it is related to xmlrpc - there are separate
>     connection, separate processes for this.
>
>     I would rather suspect the event routes - the event is in-line
>     handled in usrloc, so the save() will cover the delivery/handling
>     of the contact related events. And if you do something time
>     consuming in event_route[E_UL_CONTACT_INSERT] lets say, this will
>     reflect as processing time for the save() function.
>
>     Regards,
>
>     Bogdan-Andrei Iancu
>
>     OpenSIPS Founder and Developer
>        http://www.opensips-solutions.com
>     OpenSIPS Summit 2018
>        http://www.opensips.org/events/Summit-2018Amsterdam
>
>     On 02/05/2018 06:44 PM, Royee Tichauer wrote:
>>
>>     Bogdan,
>>
>>     Not sure I follow about breaking each function. What do you mean
>>     by 'actual internal save()'?
>>
>>     DB insert is not used.
>>
>>     We use:
>>
>>     event_route[E_UL_CONTACT_UPDATE]
>>
>>     event_route[E_UL_CONTACT_DELETE]
>>
>>     event_route[E_UL_CONTACT_INSERT]
>>
>>     btw, we also use the mi_xmlrpc_ng constantly for monitoring if
>>     that might have something to do with it (I saw a comment on
>>     opensips 1.6 that there was an opensips blocking issue in some
>>     cases).
>>
>>     Royee
>>
>>
>>
>>     On Mon, Feb 5, 2018 at 6:34 PM Bogdan-Andrei Iancu
>>     <bogdan at opensips.org <mailto:bogdan at opensips.org>> wrote:
>>
>>         Hi Royee,
>>
>>         Well, you need to break for each function the possible types
>>         of actions. Like for save():
>>             - actual internal save()
>>             - DB insert  if realtime mode used
>>             - potential events triggering - do you use any usrloc
>>         related events ?
>>
>>         Regards,
>>
>>         Bogdan-Andrei Iancu
>>
>>         OpenSIPS Founder and Developer
>>            http://www.opensips-solutions.com
>>         OpenSIPS Summit 2018
>>            http://www.opensips.org/events/Summit-2018Amsterdam
>>
>>         On 02/05/2018 06:20 PM, Royee Tichauer wrote:
>>>         Hi Bogdan, as the logs show:
>>>
>>>         save - 8912725us
>>>         t_relay - 5355333us
>>>
>>>         Let me know if you have any ideas,
>>>         Thanks,
>>>         Royee
>>>
>>>
>>>
>>>
>>>         On Mon, Feb 5, 2018 at 5:50 PM Bogdan-Andrei Iancu
>>>         <bogdan at opensips.org <mailto:bogdan at opensips.org>> wrote:
>>>
>>>             Hi Royee,
>>>
>>>             what are the time values reported for 't_relay' and
>>>             'save' ? the 50ms threashold may not be so relevant for
>>>             explaining delays of seconds.
>>>
>>>             Best regards,
>>>
>>>             Bogdan-Andrei Iancu
>>>
>>>             OpenSIPS Founder and Developer
>>>                http://www.opensips-solutions.com
>>>             OpenSIPS Summit 2018
>>>                http://www.opensips.org/events/Summit-2018Amsterdam
>>>
>>>             On 02/05/2018 02:09 PM, Royee Tichauer via Users wrote:
>>>>             Hi,
>>>>
>>>>             We are using opensips 2.1 as an SBC component and we
>>>>             see that we have loss of registrations happening once
>>>>             in a while. When we captured SIP we saw that some
>>>>             messages are being delayed by a couple of seconds. We
>>>>             enabled 'exec_execute_message=50000' and saw in the
>>>>             logs that opensips core methods like 't_realy' and
>>>>             'save' . I looked at SAR during this time and there are
>>>>             no indications of CPU running high during this time.
>>>>
>>>>             Can you help me figure out this issue?
>>>>
>>>>             Thanks,
>>>>             Royee
>>>>
>>>>             Logs:
>>>>

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.opensips.org/pipermail/users/attachments/20180206/c103d2ec/attachment.html>


More information about the Users mailing list