[Fusioninventory-user] agent control logic

[Guillaume Rousse]

Le 07/08/2013 16:46, Marco Gaiarin a écrit :
>> * new fusioninventory-inventory executable, dedicated to inventory task
>
> mean that if i need to run the agen by hand, only when needed, i
> can/must run this script?
Currently, "running the agent" means "executing each available task for 
each configured targets". Even if most of those executions immediatly 
abort for lack of control directive (see the traces in the log) that's 
still useless if you're only interested in a specific task.

The goal is to get rid of the 'one single binary for all use cases', and 
to explicitely separate different part of control logic:
- when does the agent execute (scheduling control)
- what does the agent execute (tasks list control)
- how does the agent execute (task-specific execution control)
- where does it send execution results (output control)

If just need to execute a daily local inventory, using localy-defined 
options, and send the result to a central server, executing 
'fusioninventory-inventory' from a cron task is more logical than 
executing a program than first connect to a server to check what it is 
supposed to do...
-- 
Guillaume