[IPython-User] Problems shutting down ipengines from ipcluster --sshx clusterfile

Fernando Perez fperez.net@gmail....
Tue Sep 7 01:53:15 CDT 2010


On Mon, Aug 9, 2010 at 2:35 PM, Reckoner <reckoner@gmail.com> wrote:
> The problem is that when I shut down the ipcluster command using Ctrl+C, the
> ipengines   started using the above script persist and have to be manually
> shut down.
> Is there something I'm missing here?

I'm afraid it's simply a (quite important) missing piece of
functionality, I'm sorry to say: tracking down remote engines for
cleanup is not something we're doing well at the moment.  If you run
on top of a queuing system, the job scheduler does that for you, but
we simply haven't implemented that logic ourselves.

Back when I was using our engines a lot over SSH, I kept these two
little scripts handy.  Along with a helper script that would let me
send a single ssh command to all my hosts (there's many of those on
the net, but I can give you mine if you need), these would manually
just hunt down stray controllers and engines.

amirbar[bin]> cat ipconkill

set a=`ps aux | grep ipcontroller | grep -v 'grep' | awk '{print $2}' | xargs`
echo "Controllers to kill: $a"
kill -9 $a
ps aux | grep ipcontroller | grep -v 'grep'


amirbar[bin]> cat ipengkill

set a=`ps aux | grep ipengine | grep -v 'grep' | awk '{print $2}' | xargs`
echo "Engines to kill: $a"
kill -9 $a
ps aux | grep ipengine | grep -v 'grep'


Sorry not to have a more elegant solution...



More information about the IPython-User mailing list