starcluster AMI that can boot g2.2xlarge on AWS?

From: Silverstein <no email>
Date: Thu, 22 May 2014 17:49:51 -0700

Hi,

I'm looking for a starcluster AMI that preferably has cuda built-in and
that can boot on a g2.xlarge on AWS. Being able to boot on g2.xlarge is
the primary requirement.

I found the following AMI: starcluster-cuda-small-ami (ami-580b6e68).
However, when I started it via starcluster it hung at the ssh
configuration stage:

>>> Waiting for all nodes to be in a 'running' state...

1/1 ||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||||
100%

>>> Waiting for SSH to come up on all nodes...

0/1 | | 0%


After figuring out how to log in to the node that was now running (since
the normal starcluster accounts were not accessible) I discovered it
cranking away on the GPUrunning jhProtominer. Apparently this is some
sort of bitcoin miner trojan.

This was surprising as I wasn't expecting anything to be running let
alone something like this. It was also disappointing because it means
that I can't use this AMI. And it does make me now suspicious of AMIs
including starcluster AMIs (maybe I should have been more so all along).

So my questions are:
1. where can I find a starcluster AMI that can boot on a g2.xlarge (I
tried the two HVM AMIs' listed by 'starclsuter listpublic' --
ami-486afe78 and ami-80bedfb0 -- but they both failed to boot with:
ERROR - UnsupportedOperation: Instances of type 'g2.2xlarge' may not be
launched from AMI.)
2. how are the AMIs available on the stcluster site vetted? How can I
ensure that these are safe?
3. If there is no way to verify that any given starcluster AMI is safe,
how do I build a g2.xlarge bootable starcluster AMI as there seem to be
some specific requirements for this sort of instance.

Thanks,

Herc
Received on Thu May 22 2014 - 20:49:55 EDT