[Fusioninventory-devel] Deployment in unattended-gui (and others)

[Mario Gzuk]

>> It should do something like this:
>> <?xml version="1.0" encoding="UTF-8"?>
>> <REPLY>
> I think the header should be more like that if we follow the INVENTORY XML
> format example:
> <REQUEST>
>   <CONTENT>
>   <QUERY>RUNCOMMAND</QUERY>
>> 	<RUNTASK>INSTALLATION | UPDATE | USERINSTALL</RUNTASK>
> Can't this be passed in the OPTS line too?

Sure... See later

> 
>> 	<COMMAND>wemerge.cmd | wemerge.sh</RUNTASK>
>> 	<OPTS>--automatic-service | --automatic-service-user ....</OPTS>
>> 	<ENVIRONMENT>
>> 		<VAR>FOO=BAR</VAR>
> I suggest this instead which is a bit more in the spirit of what
> already in the inventory.
>  	<ENVS>
>                 <NAME>FOO</NAME>
> 		<VAR>BAR</VAR>
>  	</ENVS>
>  	<ENVS>
>                 <NAME>FOO2</NAME>
> 		<VAR>BAR2</VAR>
>  	</ENVS>

This is ok. (but may be not needed)


> 
>> 	</ENVIRONMENT>
>> 	<SHAREDATA>
>> 		<MOUNTCMD>net use</MOUNTCMD>
>> 		<SHARE>\\server\share</SHARE>
>> 		<SHAREUSER>username</SHAREUSER>
>> 		<SHAREPASS>password</SHAREPASS>
>> 	</SHAREDATA>
> How do you pass this to you agent, is it a parameter? If so
> can't we put that in <OPTS> too?
>> </REPLY>


This might be helpful for the agent. But if you dont want to do the
mount stuff from the agent, there will be an other procedure to run the
script -> like in wpkg -> see the comment later

> 
> For wpkg, we need to be able to retrieve files from the server. I suggest
> something similar with ENVS:
> <FILES>
> <URL>http://toto/aa | \\server\aa\tt</URL>
> <NAME></NAME>
> <EXTRACT>1 | 0</EXTRACT> should the file be extracted (optional)
> </FILES>

This should be also possible with uranos, download the master control
script which makes all the stuff needed (mount, install etc...) from an
<URL> and run this with the <OPTS>. Also the <COMMAND> and the
<ENVIRONMENT> might be optional. There are a lot of different ways how
this can be solved. I will adapt the easiest solution ;)


> 
> What is the format of the answer to expect? I suggest:
> 
> <RUN_DURATION>1234</RUN_DURATION> The time used to run the command (optional)
> <RET_CODE></RET_CODE> (optional)
> <LOG> (optional)
> balbablabla
> blaglablaba
> balba
> bablabal
> </LOG>

It would be nice if the agent catch all the output from the smaster
controll script (see above) and put it into the <LOG> section.
This will be fine for me...

Following the http://en.wikipedia.org/wiki/KISS_principle:
The agent should do only main tasks so that you never have to change it:
1.) get the script from somewhere (http[s] with or whithout auth) and
store it local
2.) run the scritp with the <OPTS>
3.) catch the return code and the output from the script and
4.) send it back to the server

What do u think?

>>> A question regarding unattended-gui, is it possible to reuse the PHP server side
>>> part for example, in GLPI? Or maybe through webservice?
>>
>> What do you mean with this? The software installation part? Can you
>> explain this? What you want to do? As it is open source you can reuse
>> the code as you want. If you need a interface (SOAP, PLAIN or whatever)
>> it should be possible to integrate this into unattended-gui.
> Yes, is it possible to interact with your agent from GLPI / GOSA / Foobar?
> If so, how? I've no plan regarding that. I'm just curious to know how
> it's possible to have interaction between our different projects.
> 

What agent do you mean? Or you mean the server? In which way you think
about an interaction?

Cheers and Thank you mario