Resizing your cluster
Your HPCBOX Cluster can be resized manually by adding more nodes of a particular type using the HPCBOX ClusterApp's ClusterControl pipeline step.
The Cluster Control pipeline step is a very important part of the intelligent workflow capabilities offered by HPCBOX. You can include this step to scale-out/scale-in/start/stop your cluster workers as part of your workflow pipeline. Use of this step ensures that you optimize the expenditure on worker nodes and exercise fine grained control on your use of cluster nodes.
The START action lets you start a certain number of nodes of a certain type.
Note that, if the total number of powered down nodes is less than the count provided to START, all nodes are started.
The START_RDMA action performs the same action as START, in addition, it runs an RDMA test to verify the InfiniBand network after the nodes are started. It is highly recommended to use this method when starting compute workers which are on an RDMA network.
This option works only when your HPCBOX cluster has InfiniBand powered nodes like on Microsoft Azure.