Re: Anaconda send signal to running Dogtail testcase

[James Laska <jlaska redhat com>]

On Thu, 2007-05-10 at 14:29 +0200, Alexander Todorov wrote:
> Hello list,
> Do you think it is a good thing anaconda to inform a running dogtail 
> testcase before it exists? We can send SIGUSR1 to the dogtail testcase 
> and let it handle this.
> 
> An example what this can be used for:
> write output to dogtail logs and use %post script in kickstart to save 
> the logs on the installed machine.

Good thinking Alexander! 

> Some questions come in mind:
> 
> 1) Do we have to wait for the running dogtail process to exit normally 
> before continuing to %post step in anaconda? If we wait and the other 
> process hangs up?
> 
> The solution that I see here is send SIGUSR1 first, wait a while (how 
> long) then send SIGKILL to avoid hanging up.
> 
> 2) What is an exception in anaconda occurs? Say while handling 
> partitioning. Does it execute %post if present or directly shuts down?

I believe we can instrument logic into %traceback for the above
scenario.

> 3) The best place to implement this? I think adding a new step in 
> dispatch.py and handling exceptions is enough. What's your opinion?

One thought on perhaps a least-invasive approach (or an interim
solution) might be to have a %post in kickstart to signal a running
dogtail script to shutdown?  That will at least help us iron out the
interaction details before pushing content into anaconda?

What are your thoughts?

Thanks,
James