Command-Line Interface (CLI)

dask-mpi

Usage

dask-mpi [OPTIONS] [SCHEDULER_ADDRESS]

Options

--scheduler-file <scheduler_file>

Filename to JSON encoded scheduler information.

--scheduler-port <scheduler_port>

Specify scheduler port number. Defaults to random.

--scheduler-rank <scheduler_rank>

The MPI rank on which the scheduler will launch. Defaults to 0.

--interface <interface>

Network interface like ‘eth0’ or ‘ib0’

--protocol <protocol>

Network protocol to use like TCP

--nthreads <nthreads>

Number of threads per worker.

--memory-limit <memory_limit>

Number of bytes before spilling data to disk. This can be an integer (nbytes) float (fraction of total memory) or ‘auto’

--local-directory <local_directory>

Directory to place worker files

--scheduler, --no-scheduler

Whether or not to include a scheduler. Use –no-scheduler to increase an existing dask cluster

--nanny, --no-nanny

Start workers in nanny process for management (deprecated use –worker-class instead)

--exclusive-workers, --inclusive-workers

Whether to force workers to run on unoccupied MPI ranks. If false, then a worker will be launched on the same rank as the scheduler.

--worker-class <worker_class>

Class to use when creating workers

--worker-options <worker_options>

JSON serialised dict of options to pass to workers

--dashboard, --no-dashboard

Launch the Dashboard [default: –dashboard]

--dashboard-address <dashboard_address>

Address for visual diagnostics dashboard

--name <name>

Name prefix for each worker, to which dask-mpi appends -<worker_rank>.

Arguments

SCHEDULER_ADDRESS

Optional argument