Deployment of AWS cluster with custom AMIs  
(iii – launching the cluster and connecting to it)

Once the scrip file is ready, copy it to your AWS ParallelCluster base directory and start the cluster with $ pcluster create wrfcluster. The launch will go through several stages, as reported on the screen. This is not required but you can further visualize progress on the AWS console. For example, the following screenshot shows the on-going process while launching a small cluster with the master instance and 4 compute instances.


These are names given by AWS ParallelCluster. Do not log in onto the master instance until the process is complete even if the console shows all the instances running. The screen will indicate completion with a message like:
        Status: parallelcluster-c1cluster – CREATE_COMPLETE
        ClusterUser: centos

You can connect to the master instance and perform a preliminary check that everything is up and running with $ sinfo, which should show information about the compute instances.
When you are finished with your computations, end the cluster with $ pcluster delete wrfcluster. Manually deleting cluster resources is feasible but strongly not recommended; it is easy to forget some of the many resources used by AWS ParallelCluster and you will still need to run $ pcluster delete wrfcluster before launching any other cluster.

Close Menu