Hello Henning,

 

 

This is not a race condition.

The issue I’m observing is the following :

 

Kamailio receives a signal SIGTERM (15) when it’s being stopped.

The main process propagates the signal to the children.

 

The children terminate without closing their opened file handles. Consequently, the files are not flushed to disk.

This entails we’re losing all data written to these files since the previous automatic flush.

 

 

Did anyone try using parameter « flush = 0 » of module db_flatstore ? (the default is 1)

 

 

 

Regards,

Nicolas.

 

 

 

Hello Henning,

 

 

Thanks for your reply !

 

For the record, I’m using Kamailio module "flatstore" with the flush deactivated. In this situation, when Kamailio is stopped, I’m losing the remaining data in files not yet flushed.

So my idea was, when destroy is called, to iterate over the connection pools, and flush/close all the opened files properly.

 

But I found out that it’s not that simple :/

 

Do you think this is something that could be fixed in Kamailio ?

 

 

Regards,

Nicolas.

 

 

Hi Nicolas,

 

you are right, there is no special function for children destroy. Most modules do not need it, as the data is in shared memory, processes are started over standard kamailio APIs which will be stopped from the core automatically, to give some background.

 

Some modules do it by just iterating over an internal list of their children and then do the cleanup task. You can find an example in cdp module (destroy, diameter_peer_destroy).

 

Cheers,

 

Henning

 

 

 

Hello,

 

 

I have a module, for which I need to perform some processing in the children processes when Kamailio is being stopped.

 

I’ve looked at the "struct module_export", in which I see "mod_init", "child_init", "mod_destroy"… but no "child_destroy". :/

Is there a way to achieve this ?

 

 

Thanks for your help.

 

Regards,

Nicolas.