[SR-Users] strange --dialog in delete state is too old-- log line managing dialog hashes

Tue Nov 14 08:24:01 CET 2017

hello DAniel

are these the patches you refer to?

a3d3ea837b0f4011a3ca0e51ad4a1516bb211299
[5.0_produccion e577807] dialog: parse all headers before getting the 
message attributes for dialog

0725f750e1981be8e2058ce1a48e90eb7862e8bd
[5.0_produccion 4ff26d1] dialog: set the end time stamp immediately 
after going into deleted state

best regards
david

El 13/11/17 a las 09:42, Daniel-Constantin Mierla escribió:
> Hello,
>
> I pushed two patches to dialog module few days ago, can you try with
> that version (or with those patches backported to your version)?
>
> Cheers,
> Daniel
>
> On 09.11.17 18:50, David Escartín wrote:
>> hello Daniel
>>
>> I might think they were not, at least i cannot see dialogs being
>> created with the same id and entry hash values in the previous hours
>> to the kamailio reset, but i'm not sure. I dont know if maybe they
>> could be some "lost" dialogs with more than 2 hours lifetime for some
>> strange reason, but i could not check the dailog profile list live
>> because i got a "reply too big" output using the kamcmd rpc command.
>> I guess i should increase  binrpc_max_body_size ?
>>
>> I can try to remove traffic from the server and see if after 2 hours i
>> have any remaining dialog. I don't think there would be any because i
>> checked generally other times that the number of dialogs go to 0 after
>> a while (2 hours or so) when i remove traffic from a kamailio
>> instance. But i can check it anyway.
>>
>> i will let you know
>>
>> best regards
>> david
>>
>>
>>
>> El 08/11/17 a las 08:40, Daniel-Constantin Mierla escribió:
>>> Hello,
>>>
>>>
>>> On 08.11.17 07:23, David Escartín wrote:
>>>> Hello Daniel
>>>>
>>>> sorry about that.
>>> no worries, it was more for the future to keep a conversation in a
>>> single place, if it is not some generic announcement or similar ...
>>>> yes, if we make a restart, after a while (not fixed time some times
>>>> minutes, some times 2 hours),  we start to see those types of messages
>>> Do you know if all these dialogs were active at the last restart? Or new
>>> dialogs after restart expose the same issue?
>>>
>>> Cheers,
>>> Daniel
>>>
>>>> i attach you the sip messages of the call of the logs in the first mail
>>>> the INVITE receiver is the Kamailio instance.
>>>>
>>>> thanks a lot and sorry again about the 2 email accounts
>>>> david
>>>>
>>>>
>>>> El 07/11/17 a las 18:40, Daniel-Constantin Mierla escribió:
>>>>> Hello,
>>>>>
>>>>> first: no need to post on both sr-users and sr-dev, it makes it
>>>>> hard to
>>>>> follow up if people answer on different lists.
>>>>>
>>>>> If it is about a stable release, you can use the sr-users, if it is
>>>>> about devel version, you can use sr-dev. Of course, if it is a bug,
>>>>> you
>>>>> can open an issue on:
>>>>>
>>>>>      - https://github.com/kamailio/kamailio/issues
>>>>>
>>>>> Now, back to the message itself -- have you done a recent restart
>>>>> before
>>>>> this situation is exposed? Do you capture the traffic in your network?
>>>>> If yes, can you extract the sip packets for one of these calls and
>>>>> send
>>>>> them over to me?
>>>>>
>>>>> Cheers,
>>>>> Daniel
>>>>>
>>>>>
>>>>> On 07.11.17 16:30, David Escartín wrote:
>>>>>> hello all
>>>>>>
>>>>>> recently we are seeing some weird messages handling with dialogs in
>>>>>> Kamailio version 5.0
>>>>>> we sometimes are seeing messages like
>>>>>> /usr/local/kamailio/sbin/kamailio[15372]: NOTICE: dialog
>>>>>> [dlg_hash.c:249]: dlg_clean_run(): dialog in delete state is too old
>>>>>> (0x7fa65445c850 ref 3)
>>>>>> /usr/local/kamailio/sbin/kamailio[15372]: NOTICE: dialog
>>>>>> [dlg_hash.c:235]: dlg_clean_run(): dialog in early state is too old
>>>>>> (0x7fa652d57110 ref 1)
>>>>>>
>>>>>> we increased the debug description adding some lines to the dialog
>>>>>> module code so we could track the calls of the calls that these
>>>>>> messages belong to, and we could see that those messages appeared in
>>>>>> calls just released at that moment, for example:
>>>>>>
>>>>>> <134>Nov  4 11:21:38 localhost
>>>>>> /usr/local/kamailio/sbin/kamailio[4108]: INFO: mad-localhost-1 Call
>>>>>> 97980 / Call-ID 1409565771_82382809 at 195.219.240.46: Creating dialog
>>>>>> [8043:21772] with hash id 21772 and hash entry 8043
>>>>>> <134>Nov  4 11:21:38 localhost
>>>>>> /usr/local/kamailio/sbin/kamailio[4106]: INFO: mad-localhost-1 Call
>>>>>> 97980 / Call-ID 1409565771_82382809 at 195.219.240.46: Status 100, 6610
>>>>>> <134>Nov  4 11:21:39 localhost
>>>>>> /usr/local/kamailio/sbin/kamailio[4111]: INFO: mad-localhost-1 Call
>>>>>> 97980 / Call-ID 1409565771_82382809 at 195.219.240.46: CANCEL
>>>>>> received in
>>>>>> A-Leg, relaying downstream
>>>>>> <134>Nov  4 11:21:39 localhost
>>>>>> /usr/local/kamailio/sbin/kamailio[4112]: INFO: mad-localhost-1 Call
>>>>>> 97980 / Call-ID 1409565771_82382809 at 195.219.240.46: Status 487, 6610
>>>>>> <133>Nov  4 11:21:39 localhost
>>>>>> /usr/local/kamailio/sbin/kamailio[4139]: NOTICE: dialog
>>>>>> [dlg_hash.c:251]: dlg_clean_run(): dialog in delete state is too old
>>>>>> (0x7fa0c02a6870 ref 3) with callid
>>>>>> '1409565771_82382809 at 195.219.240.46'
>>>>>> <129>Nov  4 11:21:39 mad-proxy-inout-1
>>>>>> /usr/local/kamailio/sbin/kamailio[4112]: ALERT: dialog
>>>>>> [dlg_handlers.c:1715]: dlg_run_event_route(): after event route -
>>>>>> dialog not found [8043:21772] (1/5) (0x7fa0c02a6870) with callid
>>>>>> '1409565771_82382809 at 195.219.240.46'
>>>>>>
>>>>>> we printed the dialog id and entry hash values and we can see there
>>>>>> are no other calls creating same values in the previous hours, or
>>>>>> using same memory allocation, or same callid, so it seems like there
>>>>>> was some kind of strange issue with the dialog timers....¿?
>>>>>> By the way, this is happening only few times (80-100 times) a day
>>>>>> having many thousands of calls, so it's quite difficult for us to
>>>>>> duplicate, we couldn't do it until now.
>>>>>> We also tried to use the timer_procs 0 or 1 to use a different proc
>>>>>> timer but seems the issue happens in both scenarios.
>>>>>>
>>>>>> The configuration change we made and seems it was done when these
>>>>>> messages started to appear is to use dialog event_route when ended
>>>>>> and
>>>>>> failed to do some stuff there managing some dialog variables.
>>>>>> Does ti make any sense that attempting to use those variables could
>>>>>> cause these behaviour?
>>>>>> Do you have any idea about it could be or how we can check it deeper?
>>>>>>
>>>>>> thanks a lot and regards
>>>>>> david escartin
>>>>>>
>>>>>> _______________________________________________
>>>>>> Kamailio (SER) - Users Mailing List
>>>>>> sr-users at lists.kamailio.org
>>>>>> https://lists.kamailio.org/cgi-bin/mailman/listinfo/sr-users