This chapter discusses how to verify the availability of the nodes in your cluster, how to monitor node status, how to issue commands and copy data to the compute nodes, and how to monitor and control processes. For information on running programs across the cluster, see Chapter 3 of this document.
In order to use a Scyld cluster for computation, at least one node must be available or "up". Thus, the first priority when interacting with a cluster is ascertaining the availability of nodes. Unlike traditional Beowulf clusters, Scyld ClusterWare provides rich reporting about the availability of the nodes.
You can use either the BeoStatus GUI tool or the bpstat command to determine the availability of nodes in your cluster. These tools, which can also be used to monitor node status, are described in the next section.
If fewer nodes are "up" than you think should be, or some nodes report an error, check with your Cluster Administrator.