Open xiaol opened 10 years ago
@xiaol Currently StarCluster requires root privileges on the AMI in order to properly configure things which is why you had issues with the Amazon AMI - they require you to login as ec2-user and then use sudo everywhere you need root privileges. Until StarCluster has proper sudo support or the Amazon AMIs allow enabling root login via userdata some how I'm afraid StarCluster is incompatible with the Amazon AMIs. I'm guessing you were able to get around this using the official StarCluster images but just to make sure: which AMI are you using?
Looking at your traceback it seems that you're attaching an EBS volume to /scratch with a config similar to:
[vol myvol]
volume_id=vol-999999
mount_path=/scratch
[cluster mycluster]
...
volumes = myvol
Is that correct? If so how did you go about creating the EBS volume? My guess is you forgot to format the volume before using it. If so then in the future you can have StarCluster automatically format the volume for you using the createvolume
command:
http://star.mit.edu/cluster/docs/latest/manual/volumes.html#create-and-format-a-new-ebs-volume
@jtriley There's one problem with g2 GPU instance. The problem is that it doesn't work with any AMIs other that Amazon. I've tried all Starcluster AMIs and confirmed, that all of them have same error, that current AMI can not be used with g2 instance. So it means, that one can't use Starcluster AMIs because they're not compatible with g2 and can't use Amazon AMIs since they don't allow for root user.
Amazon Linux AMI 2013.09.1 - ami-b1fe9bb0 (64-bit) has the user that's ec2-user not root. As a result, running the default cluster setup ,will waiting for the ssh done forever.
then i changed to ubuntu instance , ssh is all right, but got this
So anybody know how to resolve this?
Should i chose a different instance, or apply a patch?