Now that we have a cluster configuration defined, we can build a cluster. Before building the cluster, we need to choose a cluster name, here we will use the name “cfd”. This name is used whenever we want to interact with the cluster through the “pcluster” command. The command below is used to launch a cluster. Note: Once this command is entered you will start incurring charges for the instances.
It is imperative to follow the clean-up steps to stop incurring charges. The tutorial is designed to allow for a walk-through time of 1 hour or less, in order to limit costs. The exact costs will depend on your region but it will be less than $10 if used for less than an hour.
pcluster create cfd -c ~/.parallelcluster/config
The launch process takes about 10 minutes. During the launch process, different status messages will be printed to the terminal window. A summary of this information is displayed below.
Status: CREATE_COMPLETE MasterServer: RUNNING ClusterUser: ec2-user MasterPrivateIP: 10.0.0.22 ComputeFleetStatus: RUNNING
If there are any issues with the launch process, please check if you’ve followed all the instructions and if needed look to the Parallel Cluster documentation for help. You should also check in your console for any resources that have been created (e.g EC2 console page).
You can follow the same instructions to create an Arm-based cluster, just changing the name of the cluster and the config file
pcluster create cfdarm -c ~/.parallelcluster/configarm