Deployment of AWS cluster with custom odyhpc AMIs (i – preliminary steps)
- AWS ParallelCluster Configuration
The first step is to set up the AWS credentials if you have not done it yet. The command $ aws configure will ask for the IAM’s Access Key ID and Secret Access Key along with the default region.
The second step is to begin the configuration with the $ pcluster configure command. The system will ask you a few standard questions. We recommend answering most everything by default except when prompted about the automatic creation of a Virtual Public Cloud (VPC), which requires deciding whether to create a new VPC dedicated to clusters or use an existing VPC. If you select the former, we recommend choosing a public subnet for the head node and private subnets for the compute nodes; once the information is introduced, the system will automatically create several VPC resources that will also appear in the VPC menu of the AWS console. If the choice is to use an existing VPC and subnet, you will need their IDS as discussed in the ‘cluster design’ subsection. In either case, the command pcluster will create a subdirectory (e.g. .parallelcluster) with a config file in it. You can substitute this file a bit later, but it contains some identification parameters that will be needed during the cluster design phase.
The third step is simply to check the version with $ pcluster version (without any dash), which returns the AWS ParallelCluster version. It is strongly encouraged to match the running AWS ParallelCluster version with the version used to create the AMI. This piece of data is detailed in the awspcluster_version.txt file located at the home/centos/CLUSTERs subdirectory. AMIs created with an older pcluster version usually do not work well with a newer version of pcluster. Therefore, and if the versions do not match, we recommend changing the pcluster version with $pip/pip3 install aws-parallelcluster==[awsparallelcluster_version_used_to_create_AMI].