StarCluster TODOs

Below are the current feature requests for the StarCluster project that need implementing:

Support for Amazon VPC

Add the ability to configure StarCluster to launch the cluster instances inside of an Amazon VPC. Should just involve adding a SubnetId parameter to the run_instances boto call. Feature requested by Adam Kraut.

Improved Eucalyptus Support

Need to subclass starcluster.cluster.Cluster and starcluster.awsutils.EasyEC2 for Eucalyptus to handle the lack of the following API features in ECC:

Add More Options to “addnode” command

The following options need to be added to the addnode command:

  • –image-id (-i)
  • –instance-type (-I)
  • –availability-zone (-z)
  • –bid (-b)

Dan Yamins has a pull request for this that needs to be merged.

Add Support for Load Balancing a Given SGE Queue

Load balancer should support balancing Sun Grid Engine queues other than just all.q. This is useful if you want to load balance many different queues with varying configurations. In this case you can launch a separate load-balancer process for each queue.

Dan Yamins has a pull request for this that needs to be merged.

Use PyStun to Restrict Cluster Acccess to User’s IP-address

StarCluster should support restricting ssh access to the user’s current ip-address when creating a new cluster. This feature will need to use the pystun project to correctly determine the user’s public ip behind firewalls and NAT. StarCluster’s ssh* commands will also need to be modified to check that the user’s current ip has been allowed to access the cluster in the case that they use StarCluster from multiple machines.