[Simfactory] Test script

[Erik Schnetter]

On Mon, Mar 14, 2011 at 8:34 PM, Ian Hinder <ian.hinder at aei.mpg.de> wrote:
>
> On 2 Mar 2011, at 22:44, Erik Schnetter wrote:
>
>> SimFactory 2 (the Python version) comes with a script
>> "simfactory/bin/distribute" that tests syncing, configuring, building,
>> submitting, and running a test job to remote systems. Instructions for
>> using this script are in the script. If you are developing SimFactory
>> 2, I recommend using this script to test your changes.
>
> I've just been trying out this script, and have a few questions:
>
> 1. Is it supposed to work on a machine with no queuing system?  It would be nice to test it quickly on a laptop before trying it out on a remote cluster.

Yes it is. I always test on my local laptop. You need a submit script
for this -- I use submit scripts that just start a job in the
background.

> 2. Is it supposed to detect that a build has failed and then not try to submit a job?  At the moment it seems to try anyway.

That is a good idea. At the moment it tries anyway and then fails.

> 3. Is there an easy way to abort a test once it has been started?

Unfortunately not. The tests are started in the background (I output
process ids), and then use sim --remote to access a remote machine. I
have not found a good way to kill running tests. One extreme way is a
"killall sim; killall python; killall ssh; killall rsync" on the local
system, then log into the remote system and issue "kill -1 -1". (Don't
do this unless you know what it does...) Another way is just to wait
until activity as died.

> 4. For the purpose of testing simfactory, it would be good to default to a very quick thornlist.  I remember some discussion of an "empty.th" but I can't find that discussion or the thornlist.

I just moved it into simfactory/etc/parfiles, but haven't committed it
yet. This is an empty file.

> Thanks for the script!

You are welcome!

I worked on some improvements over the weekend; it now also tests
cleanup and presubmitting. I was about to test checkpointing, but will
now wait until the patch is in. I should probably commit the current
state.

-erik

-- 
Erik Schnetter <schnetter at cct.lsu.edu>   http://www.cct.lsu.edu/~eschnett/