[sr-dev] RFC: removing (disabling) fork=no

[Daniel-Constantin Mierla]

On 21/01/16 08:46, Olle E. Johansson wrote:
>> On 20 Jan 2016, at 09:19, Daniel-Constantin Mierla <miconda at gmail.com> wrote:
>>
>> Hello,
>>
>> wondering if anyone is using fork=no -- some old docs suggest it is
>> suitable for debugging, but actually kamailio doesn't work properly in
>> this mode, leading to more troubles than benefits (e.g., having reports
>> of invalid issues, like tcp not working in this mode).
>>
>> In first phase I would disable setting this value, with a warning if set
>> to no, because most of the configs out there have fork=yes. Removing it
>> could be considered in the future.
>>
>> Note that this fork=no is different than don't daemonize controlled with
>> -D, which will stay being useful for some init.d systems.
> What is the difference?
>
> I have been using fork=no a lot in test scripts, but could possibly move to
> -D. Having a config file parameter is easier though, said the lazy man.
kamailio -h gives some short explanation.

When fork=no, no sip workers are created, tcp not working, if no listen
is set, it binds only to first network interface (can be loopback). This
is like -D.

-DD is main process is not daemonizing itself, so stays attached to the
terminal (so can be killed with CTR+C) but the rest of processed (sip
workers, timers) are forked like normal.

-DDD is like fork=yes, the main process being optimized as well.

For proper debugging and also staying attached to terminal, the -DD and
children=1 is the best combination.

Cheers,
Daniel

-- 
Daniel-Constantin Mierla
http://twitter.com/#!/miconda - http://www.linkedin.com/in/miconda
Book: SIP Routing With Kamailio - http://www.asipto.com
http://miconda.eu