[StarCluster] SSH doesn't come up on c3.8xlarge instances

[Christian Zielinski]

Dear all,

StarCluster proved to be an extremely useful tool in my research (thanks
for everybody being involved!), but currently I'm experiencing some
strange behavior. I run all my instances in the us-east-1 region with
the AMI:

ami-6b211202 us-east-1 starcluster-base-ubuntu-13.04-x86_64-hvm (HVM-EBS)

That works very well for c3.{large, xlarge, 2xlarge, 4xlarge} instances.
However, when I start a c3.8xlarge instance with the same AMI SSH
doesn't get up.

Specifically, after starting a c3.8xlarge instance StarCluster is
hanging at this point:

  >>> Waiting for SSH to come up on all nodes...
  0/1 |                                           |   0%

In the AWS web console I can find the following message under "Instance
Status Checks" (the system status check is fine):

"Instance reachability check failed at April 23, 2014 11:08:00 AM UTC+8
(7 minutes ago)"

What really confuses me is that everything works fine on all the other
instance types and only c3.8xlarge seems to be affected. Could it be
potentially some compatibility issue of the AMI? Interestingly with the AMI

ami-52a0c53b us-east-1 starcluster-base-ubuntu-12.04-x86_64-hvm (HVM-EBS)

everything works fine. However, I would like to use the newer AMI (among
the reasons is enhanced networking and being a HVM image). Does anybody
experience similar problems?

Thank you and regards
Christian Zielinski


P.S.: The system logs are ending with the following messages:

[...]
[    0.802898]  #30cpu 31 spinlock event irq 281
[    0.818286] installing Xen timer for CPU 31
[    0.820037]  #31
[    0.835184] Brought up 32 CPUs
[    0.836006] smpboot: Total of 32 processors activated (180284.38
BogoMIPS)


-- 
http://www.czielinski.de/