Carolyn wrote:
>There was a bit of discussion earlier this summer (see 
>http://article.gmane.org/gmane.comp.voip.sipx.devel/11076 
><http://article.gmane.org/gmane.comp.voip.sipx.devel/11076> ) 
>about running allowing supervisor to run "commands" rather 
>than "services" at the request of sipXconfig (i.e. via 
>xmlrpc).  Scott mentioned creating an issue for a 
>sipXecs-command definition, but I'm not sure whether that 
>actually happened.
>
>We now have at least two examples of this: 
>
>- the package update stuff that Raymond is working on, and
>- run configtest independently which I am starting to look at 
>
>At the time of the earlier discussion it was mentioned that 
>being able to execute sipx-snapshot on a remote machine would 
>be another example.
>
>Raymond and I are wrestling with some of the same issues and 
>we have some different ideas about how to do this. 
>
>Raymond is proposing launching a command, redirecting its 
>output to a file, and providing an interface to return the 
>contents of the file.  He is wrestling with how to prevent the 
>same command from being run more than once at the same time.  
>The package update command requires sudo access, and involves 
>shutting down the entire sipxpbx.
>
>I had thought of running my command (configtest) under the 
>control of a task in supervisor, and collecting the output the 
>same way we collect stdout/stderr for the processes.  Since 
>the pid executing the command would be under the control of 
>the supervisor, it would be easier to prevent it from being 
>executed twice.  However the command wouldn't survive a 
>shutdown of sipxpbx.  (That's a good thing for me)
>
>We have to decide on how to know what to execute.  From a 
>command definition in xml similar to what the process 
>definitions use?  (that's a natural choice for me, since 
>configtest is already defined that way) or a hard-coded 
>command with just the prefix determined at install?.  
>
>Both of us are worried that we are opening up the pbx to 
>possible misuse. 
>
>Does anyone think that we should be implementing a generic 
>"command" capability, which would launch a command and return 
>its output? or should we just implement the two specific needs 
>we have, and not worry about a generic design?
>

At this stage of the development, I think we should just implement
exactly what we need.  I keep running into issues every step of the
development for the software updating so I think that trying to do a
generic command execution would be too risky.   
_______________________________________________
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