Robert Joly wrote:
>> On Fri, 2008-12-05 at 12:36 -0500, Raymond Dans wrote:
>>
>>> 1. The owners of the Processes that have IMDB tables like 
>> "caller-alias"
>>> and "userlocation" need to modify their process definition files to 
>>> mark (using the attribute required) these tables as 
>> "required=false".
>>
>> Correct.  caller-alias should be declared by sipXproxy, and 
>> userlocation by sipXregistrar, if I'm not mistaken 
>> (technically, both of these are used by plugins loaded by 
>> those components, but we have not yet dealt with that 
>> subtlety in our process management scheme, and won't for this 
>> release).
> 
> I do not agree that these tables should be declared with
> 'required=false'  the reason being that if sipXproxy and sipXregistrar
> are allowed to start without these IMDB tables being available,
> unexpected results will occur when someone either makes a call that
> involves caller aliases or tries to make a location-routed call. The
> sipXproxy and sipXregistrar components must be launched after
> caller-alias and userlocation respectively are loaded to ensure that
> these components behave as per the configuration of the box right from
> the start.
> 
> This does impose that sipXconfig initiate the replication of all IMDBs
> even when empty at start-up though to satisfy the dependency.
> 
> [...]

For what it's worth I have no problem with that requirement. sipXconfig
does push all datasets every time it starts (which is probably not correct
BTW: it should be only done after upgrade and on fresh install), but the
order in which things happen during start-up looks a bit too random to me.

It needs to be synchronized: resources should be generated before we set
the configuration timestamps and before we actually tell supervisor to
start services.
D.

_______________________________________________
sipx-dev mailing list
[email protected]
List Archive: http://list.sipfoundry.org/archive/sipx-dev
Unsubscribe: http://list.sipfoundry.org/mailman/listinfo/sipx-dev

Reply via email to