Hi Daniel,

this is the way we are creating the child processes (1 manager and n workers)
the problem is with the workers.
the worker will call a event-route and in the script config we try to call pres_refresh_watchers and that's where we get the pa_db = NULL.

as i understand from your email, if we change PROC_NOCHLDINIT and let the child_init execute for the forked process then it will also execute child_init in other modules ? it makes sense.

going to try this.


static int mod_child_init(int rank)
{
int pid;
int i;

fire_init_event(rank);

if (rank==PROC_MAIN) {
pid=fork_process(PROC_NOCHLDINIT, "AMQP Manager", 1);
if (pid<0)
return -1; /* error */
if(pid==0){
kz_amqp_manager_loop(0);
}
else {
for(i=0; i < dbk_consumer_processes; i++) {
pid=fork_process(PROC_NOCHLDINIT, "AMQP Consumer", 1);
if (pid<0)
return -1; /* error */
if(pid==0){
mod_consumer_proc(i+1);
}
}
}
}

return 0;
}



On Mon, Sep 8, 2014 at 1:11 PM, Daniel-Constantin Mierla <miconda@gmail.com> wrote:
Hello,

the database connection should not be shared beween processes, because it can bring unexpected results in may places.

Right now, the rule is to have one connection per process, shared by all modules in that process.

To achieve that, at mod_init each module opens database connection and closes it before ending the function. Then in child_init() the connection is opened again. Another module that will have to open in child_init() will get the same connection now.

When you create a new process, you tell the type of child and based on that child_init() from the other modules are executed.

What is the function do you use for creating a new process? Maybe you can paste it here exactly how you do it and I can see if something can be done.

Cheers,
Daniel


On 03/09/14 12:09, Luis Azedo wrote:
Hi Jason,

thanks for the reply.

the last 2 statements in presence module mod_init close the connection and set pa_db to NULL. when my module main process forks the extra processes the pa_db in presence is NULL, so when it calls pres_refresh_watchers it fails because pa_db is NULL.
i commented these last statements in presence mod_init and i got it to work.

// pa_dbf.close(pa_db);
// pa_db = NULL;

does this have any implications on how the module works ? is it ok to merge this change ?

Thank you



----------------------------------------------------------------------

Message: 1
Date: Sun, 31 Aug 2014 09:40:49 +0200
From: Jason Penton <jason.penton@gmail.com>
To: "Kamailio (SER) - Development Mailing List"
        <sr-dev@lists.sip-router.org>
Subject: Re: [sr-dev] accessing database from a child forked process
Message-ID:
        <CALoGXNWvHhCAO91Tfa0w8W3eYQRvfV7Qkgte7dBnD+ciNr_Kpg@mail.gmail.com>
Content-Type: text/plain; charset="utf-8"

To confirm exactly what processes are being used, maybe check the log file
and take note of process id at each log event. For example you could check
the process id of the messages showing you the connection is null. Then run
kamcmd ps to show the process list with a description of each Kamailio
process. That will probably point you in the correct direction.

Cheers
Jason


On Fri, Aug 29, 2014 at 3:53 PM, Luis Azedo <luis.azedo@factorlusitano.com>
wrote:

> Hi,
>
> i have a module that creates 1 extra process where it processes stuff in a
> loop.
> on some condition i fire a route_event with a fakemsg and its up to the
> user of the module to take action, it tries to work like dispatcher module
> or htable (mod-init) events.
>
> the problem that i have is that, if i call some function on some module
> that performs a database action, the connection is null when it calls
> use_table.
>
> in this case i'm making this call
> event_route[my_module:my_event]
> {
>  $var(my_uri) = <<result of some operations>>;
>  pres_refresh_watchers("$var(my_uri)", "dialog", 1);
> }
> presence module makes the call to use_table but this call fails because
> the connection is null. presence module is working fine besides this.
>
> if i make this call inside event_route[dispatcher:dst-up] it works.
>
> i think that dispatcher fires the event inside a callback from a
> registered timer, so, i think (may be wrong) that it comes from a different
> process ?
>
> i wonder if i'm missing something from child_init ? need to register
> something ?
>
> thanks for your help.
>
>
>
>
>
>
> _______________________________________________
> sr-dev mailing list
> sr-dev@lists.sip-router.org
> http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev
>
>
-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://lists.sip-router.org/pipermail/sr-dev/attachments/20140831/9fba51e4/attachment-0001.html>

------------------------------

_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev


End of sr-dev Digest, Vol 70, Issue 71
**************************************



_______________________________________________
sr-dev mailing list
sr-dev@lists.sip-router.org
http://lists.sip-router.org/cgi-bin/mailman/listinfo/sr-dev


-- 
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Next Kamailio Advanced Trainings 2014 - http://www.asipto.com
Sep 22-25, Berlin, Germany