Class | Description |
---|---|
MPIMasterStarter |
This class is started in the first pod in a StatefulSet
This class will get the PodIP addresses from all pods in a job
When getting the IP addresses, it also waits for all pods to become running
It saves those IP addresses to hostfile
It checks whether password free ssh is enabled between this pod and
all other pods in the statefulset
It then executes mpirun command to start OpenMPI workers
|
MPIWorkerStarter |