Kubernetes之AllReduce任务调度 (CPU+GPU环境) --【D-8/14】Open MPI集群之Hello World测试

Abstract

The Open MPI cluster is ready. This blog will test the Open MPI cluster on multiple hosts.

Note that an assumption is that the remote NFS file-sharing directory is mounted on the same local directory. In other words, for convenience purposes, the $HOME/Public/nfs1 at NFS server is mounted on $HOME/Public/nfs1 at NFS clients.

1. Preliminaries

Our previous blog showed how a master node sends commands to workers’ nodes without manual interaction of password inputs.

Also, our previous blog showed how accessing a source file via the network among the master and the workers.

Now, the master is going to send commands to the workers, and both the master and the workers will compile and execute the same source file. All commands and files are only operated on the master’s side, no manual interaction between master and worker is required.