Hello,

can you grab and send the backtrace with gdb from corefile? The logs show that core was generated. While I don't use dmq that much, maybe it is something easy to fix.

Cheers,
Daniel

On 18.10.17 22:58, : Paolo Visintin - Time-Net S.r.l. wrote:
Hi Daniel, 
I have gone in deep and made more tests, the behaviour is this: 

kamailio.01 and kamailio.02 shares the same dialog database, for sync purpose.
If call is handled by kamailio.01 I can see dialogs only in kamailio.01 (kamcmd dlg.list)
If call is handled by kamailio.01 and, in the meantime, I restart kamailio.02 I can see dialogs in both!

So I suppose that there's a sync memory > DB and not a bi-directional sync. Only at startup kamailio fetch existing dialogs and store them in memory, right ?

I tried another approach now, enabled DMQ sync with: 
modparam("dialog", "enable_dmq", 1)
with separate port dedicated for DMQ messages

Dialogs are synced, but when a call is hangupped kamailio crashes! 
4(1570) INFO: tmx [t_var.c:527]: pv_get_tm_reply_code(): unsupported route_type 64 - code set to 0
22(1588) CRITICAL: <core> [core/pass_fd.c:277]: receive_fd(): EOF on 15
 0(1549) ALERT: <core> [main.c:742]: handle_sigs(): child process 1570 exited by a signal 11
 0(1549) ALERT: <core> [main.c:745]: handle_sigs(): core was generated
 0(1549) INFO: <core> [main.c:768]: handle_sigs(): terminating due to SIGCHLD
 5(1571) INFO: <core> [main.c:823]: sig_usr(): signal 15 received
...
 1(1567) INFO: <core> [main.c:823]: sig_usr(): signal 15 received
 0(1549) CRITICAL: <core> [main.c:649]: sig_alarm_abort(): shutdown timeout triggered, dying

My purpose is to share dialogs and usrloc data between kamailio instances (10 about) in order to manage shared dialogs.

Cheers,
Paolo


2017-10-18 18:06 GMT+02:00 Daniel-Constantin Mierla <miconda@gmail.com>:

Hello,


On 17.10.17 10:27, : Paolo Visintin - Time-Net S.r.l. wrote:
Hi kamailio users,  I'm wondering if there's something already done with shared dialog among more kamailio instances, in order to manage branches initalized by another kamailio instance.

​A
ctually seems not possible (also with db mode) because kamailio is looking into caller/callee_sock and if it's different does not manage

in a "standard" failover environment (master / slave) with vIP and keepalived and $fs = VIRTUAL_IP everything is working fine

but in a distributed environment this could not happen as the relay is managed by a kamailio proxy


I looked at the code and the dialog module just not sets the local socket fields if there is no match, but loads them from db and all should be fine, an existing socket will be used if there is a need to send BYE. What exactly you encointered? There is a warning message when the local socket is not matched, but it's about ignoring the socket field, not ignoring the dialog from db.

Cheers,
Daniel
-- 
Daniel-Constantin Mierla
www.twitter.com/miconda -- www.linkedin.com/in/miconda
Kamailio Advanced Training, Nov 13-15, 2017, in Berlin - www.asipto.com
Kamailio World Conference - www.kamailioworld.com


-- 
Daniel-Constantin Mierla
www.twitter.com/miconda -- www.linkedin.com/in/miconda
Kamailio Advanced Training, Nov 13-15, 2017, in Berlin - www.asipto.com
Kamailio World Conference - www.kamailioworld.com