[IPython-User] Parallel over ssh and shared NFS directories with 0.13 on linux

Jose Gomez-Dans jgomezdans@gmail....
Mon Aug 13 10:51:37 CDT 2012


On 10 August 2012 20:31, MinRK <benjaminrk@gmail.com> wrote:

> Let me start by saying that in simple NFS+SSH environments (which I use
> most often), I find ipcluster more trouble than it's worth, and tend to do
> the following, which requires zero config:
> [host0] $> ipcontroller --ip=* &
> [host0] $> ssh host1 ipcluster engines -n 4

OK, this is good enough for the time being, but it'd be really nice to be
able to kill things cleanly. I find that otherwise, the slaves don't
recognise changes in the code (ie they need a reload). At the moment I
succeed by killing ssh sessions, not very elegant ;-)

> This is yet another bug in ipcluster argument relaying, but if you change
> SSHEngineSetLauncher.remote_profile_dir to
> SSHEngineLauncher.remote_profile_dir, I think it should work.  What you are
> doing definitely *should* work, and I will look into why it doesn't.

Nope, doesn't work. I can't retest, as I'm actually using the cluster as
above, but is there anything you'd like to know about my setup before I try

-------------- next part --------------
An HTML attachment was scrubbed...
URL: http://mail.scipy.org/pipermail/ipython-user/attachments/20120813/f699a726/attachment.html 

More information about the IPython-User mailing list