12. Launching applications

PRRTE can launch processes in a wide variety of environments, but they can generally be broken down into two categories:

  1. Scheduled environments: these are systems where a resource manager and/or scheduler are used to control access to the compute nodes. Popular resource managers include Slurm, PBS/Pro/Torque, and LSF.

  2. Non-scheduled environments: these are systems where resource managers are not used. Launches are typically local (e.g., on a single laptop or workstation) or via ssh (e.g., across a small number of nodes).

PRRTE provides two commands for starting applications:

  1. prun - submits the specified application to an existing persistent DVM for execution. The DVM continues execution once the application has completed. The prun command will remain active until the application completes. All application and error output will flow through prun. prun must run on a node where one of the DVM’s daemons is running: see Using prun with a running DVM.

  2. prterun - starts a DVM instance and submits the specified application to it for execution. The DVM is terminated once the application completes. All application and error output will flow through prterun.

The rest of this section usually refers only to prterun, even though the same discussions also apply to prun because the command line syntax is identical.

Note

“Until the application completes” includes anything the application spawns. A job started with PMIx_Spawn (or MPI_Comm_spawn) outlives the job that started it by default, and both commands wait for the whole spawn tree: output from every job in the tree is forwarded, and the exit status reported is the first non-zero status returned by any of them — the primary job or anything descended from it. Use the report-child-jobs-separately runtime option (prterun --help runtime-options) to have only the primary job’s status decide the command’s exit status, with a spawned job’s non-zero status reported on its own instead.

A job a tool starts is not in that tree. A prun run from a job script launches a step, not a child of the script: prun waits for the step, and the step’s output and exit status are returned to it, for the script to act on.