Carolyn Beeton wrote:
>> -----Original Message-----
>> From: Lawrence, Scott (BL60:9D30) 
>> Sent: Monday, September 15, 2008 1:49 PM
>> To: Beeton, Carolyn (CAR:9D60)
>> Cc: [EMAIL PROTECTED]
>> Subject: Re: [sipX-dev] process management restructure question
>>
>>
>> On Mon, 2008-09-15 at 11:08 -0400, Carolyn Beeton wrote:
>>> Is the existing RPC defined in ProcMgmtRpc still intended 
>> to function 
>>> as currently implemented, but sending requests on the the new 
>>> sipXsupervisor/SipxProcess?  (this would be 
>> setUserRequestState[All] 
>>> where state is one of USER_PROCESS_STOP, USER_PROCESS_START, or
>>> USER_PROCESS_RESTART)
>>>
>> My assumption so far has been that we would try to minimize 
>> what sipXconfig needs to change to be compatible with the new 
>> sipXsupervisor process management, and translate the "old" 
>> rpc commands into the new process state model events 
>> internally.  I had not gotten far enough to be sure this 
>> would not cause problems, though.  
>>
>> The old process model was that sipXconfig (or some other 
>> xml-rpc caller, in theory) issued stop/start/restart 
>> commands.  The new one is really that start and stop become 
>> changes to the desired state, and restart becomes, in effect, 
>> "stop this now but don't change the desired state, so it will 
>> start again right away".  Whether or not we should change the 
>> xml-rpc interface may be worth discussing, but my feeling has 
>> been that we could probably avoid it.
>>
>> The one important differenc that I think we will want to 
>> expose is that the new supervisor process model has a lot of 
>> information about _why_ a process is not starting, and that 
>> should be made available to sipXconfig somehow so that it can 
>> either act on it or display it for the admin.
>>
>>
>>
> 
> The existing rpc has a "blockForStateChange" parameter.  Am I correct to
> assume that this goes away, replaced by each process's knowledge of its
> dependencies, based on the process definition? (It's a bad idea for
> xml-rpc stuff to block anyway, as the underlying connections time
> out...)
> 

I am not clear how the "blockForStateChange" is related to dependencies
between the processes. I was under impression that this is a way of making
synchronous vs. asynchronous calls.
If I remember correctly this is the only way sipXconfig is using this API
at the moment (all those calls are issued as blocking calls).
Why is it a bad idea that RPC call is blocking? What exactly times-out? If
something does time out isn't it an indication of failure anyway?

> Another question: how do the processes get "enabled" in the first place?
> They are currently coming up "disabled".  Are we expecting sipXconfig to
> send a start command (xmlrpc) for each process on system start up
> (reflecting the HA distribution of services)? 
> 
> Carolyn

_______________________________________________
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