Hi Alex,
sounds strange indeed. 4.3.5 is from March 2016, a lot happened since then..
One idea to investigate further - have you tried already to get more debugging information from dialog? With debugger module you could restrict it to dialog, if this is a loaded production instance.
Another idea - are you using functionality like timeout_avp or sst module to update/modify the dialog timeout?
Cheers,
Henning
Am 27.08.19 um 22:31 schrieb Alex Balashov:
Hi,
I am running 4.3.5:1b0c0a, which I am aware is an EOL'd release train, and have a problem with the dialog module I am baffled by.
On many calls - I can't find any correlation to particular kinds of endpoints - I see spoofed BYEs come out of Kamailio after a minute and a half, as if the call had hit a dialog timeout or the dialog module's dead peer detection (OPTIONS pinging) kicked it out.
The thing is, neither of those explanations seem to be borne out by the parameters of the calls under investigation. The dialog timeout on these calls is set to 7200, and the dialog keepalives aren't enabled in either direction. If they were, it'd be logged.
Calls that are terminated in this manner also see the following log message:
WARNING: dialog [dlg_req_within.c:214]: bye_reply_cb(): inconsitent dlg timer data on dlg 0x7f9997317e48 [3390:9644] with clid '1996679936_133218050@x.x.x.x' and tags '2bb17663-co4006-INS001' 'as7925780d'
But I don't think this is unusual for a dialog-spoofed BYE; presumably this is due to a 200 OK for the BYE from both ends.
So anyway, I am trying to track down anything else that could inadvertently cause dialog to hang up calls relatively quickly in that release, or inadvertently set the dialog timeout parameter to something lower than is apparent from the logging and the config.
If anyone has any pointers, that would not go unappreciated!
Many thanks,
-- Alex