Hi Robert,
You guessed it. I recently discovered this same problem after trying to create
a new 9.10 karmic AMI for StarCluster a couple days ago. Luckily I was simply
using the wrong base AMI's and not the ones he published on alestic.com. I
made the mistake, unfortunately, of simply picking an AMI from ElasticFox. In
any event, with his latest 9.10 AMI's, this problem seems to have gone away.
For the short term, yes modifying that script in the current StarCluster
public AMI should fix the issues you and others have been having with
rebundled instances not coming up. The reason that script keeps StarCluster
from running a rebundled cluster is that StarCluster relies on ssh being fully
functional to do it's magic and without that /etc/rcS.d/S50ec2-ssh-host-key-
gen script executing ssh doesn't work.
I'm actually in the process of creating a new AMI as I type this email.
I'll make it public as soon as I can successfully launch a cluster. I will
also be posting my "recipe" for building this new StarCluster AMI from the
ground up in the starcluster documentation.
I'm testing this new ami with the latest version of StarCluster on github,
however, it should in theory still work for the 0.90 version. For those of you
that are interested in helping me test StarCluster, please have a look at
installing the latest development version:
http://web.mit.edu/stardev/cluster/documentation/installation.html
Thanks for bringing this up and hope that helps,
~Justin
Received on Tue Feb 09 2010 - 12:02:22 EST