slurm_launcher#

Provides a module for launching Fluent within a Slurm environment.

Examples#

Note that the keys scheduler_headnode, scheduler_queue and scheduler_account are optional and should be specified in a similar manner to Fluent’s scheduler options.

>>> slurm = pyfluent.launch_fluent(
...   scheduler_options={
...     "scheduler": "slurm",
...     "scheduler_headnode": "<headnode>",
...     "scheduler_queue": "<queue>",
...     "scheduler_account": "<account>"
...   },
...   additional_arguments="-t16 -cnf=m1:8,m2:8",
... )
>>> type(slurm)
<class 'ansys.fluent.core.launcher.slurm_launcher.SlurmFuture'>
>>> slurm.pending(), slurm.running(), slurm.done() # before Fluent is launched
(True, False, False)
>>> slurm.pending(), slurm.running(), slurm.done() # after Fluent is launched
(False, True, False)
>>> session = slurm.result()
>>> type(session)
<class 'ansys.fluent.core.session_solver.Solver'>
>>> session.exit()
>>> slurm.pending(), slurm.running(), slurm.done()
(False, False, True)

# Callable slurm launcher

>>> from ansys.fluent.core.launcher.launcher import create_launcher
>>> from ansys.fluent.core.launcher.pyfluent_enums import LaunchMode, FluentMode

>>> slurm_meshing_launcher = create_launcher(LaunchMode.SLURM, mode=FluentMode.MESHING)
>>> slurm_meshing_session = slurm_meshing_launcher()

>>> slurm_solver_launcher = create_launcher(LaunchMode.SLURM)
>>> slurm_solver_session = slurm_solver_launcher()

Classes:

`SlurmFuture`(future, job_id)	Encapsulates asynchronous launch of Fluent within a Slurm environment.
`SlurmLauncher`([mode, ui_mode, ...])	Instantiates Fluent session within a Slurm environment.

class ansys.fluent.core.launcher.slurm_launcher.SlurmFuture(future, job_id)#

Bases: object

Encapsulates asynchronous launch of Fluent within a Slurm environment.

The interface is similar to Python’s future object.

Methods:

`__init__`(future, job_id)	Initialize SlurmFuture.
`add_done_callback`(fn)	Attaches the callable function.
`cancel`([timeout])	Attempt to cancel the Fluent launch within timeout seconds.
`done`()	Return `True` if the Fluent launch was successfully cancelled or Fluent was finished running, otherwise `False`.
`exception`([timeout])	Return the exception raised by the Fluent launch.
`pending`()	Return `True` if the Fluent launch is currently waiting for Slurm allocation or Fluent is being launched, otherwise `False`.
`result`([timeout])	Return the session instance corresponding to the Fluent launch.
`running`()	Return `True` if Fluent is currently running, otherwise `False`.

__init__(future, job_id)#: Initialize SlurmFuture.

add_done_callback(fn)#

Attaches the callable function. The function will be called, with the session as its only argument, when Fluent is launched.

Parameters:

fnCallable: Callback function.

cancel(timeout=60)#

Attempt to cancel the Fluent launch within timeout seconds.

Parameters:

timeoutint, optional: timeout in seconds, by default 60

Returns:

bool: True if the Fluent launch is successfully cancelled, otherwise False.

done()#: Return True if the Fluent launch was successfully cancelled or Fluent was finished running, otherwise False.

exception(timeout=None)#

Return the exception raised by the Fluent launch. If Fluent hasn’t yet launched, then this method will wait up to timeout seconds. If Fluent hasn’t launched in timeout seconds, then a TimeoutError will be raised. If timeout is not specified or None, there is no limit to the wait time.

If the Fluent launch completed without raising, None is returned.

Parameters:

timeoutint, optional: timeout in seconds

Returns:

Exception: The exception raised by the Fluent launch.

pending()#: Return True if the Fluent launch is currently waiting for Slurm allocation or Fluent is being launched, otherwise False.

result(timeout=None)#

Return the session instance corresponding to the Fluent launch. If Fluent hasn’t yet launched, then this method will wait up to timeout seconds. If Fluent hasn’t launched in timeout seconds, then a TimeoutError will be raised. If timeout is not specified or None, there is no limit to the wait time.

If the Fluent launch raised an exception, this method will raise the same exception.

Parameters:

timeoutint, optional: timeout in seconds

Returns:

Union [Meshing, PureMeshing, Solver, SolverIcing]: The session instance corresponding to the Fluent launch.

running()#: Return True if Fluent is currently running, otherwise False.

class ansys.fluent.core.launcher.slurm_launcher.SlurmLauncher(mode=None, ui_mode=None, graphics_driver=None, product_version=None, dimension=None, precision=None, processor_count=None, journal_file_names=None, start_timeout=-1, additional_arguments='', env=None, cleanup_on_exit=True, start_transcript=True, case_file_name=None, case_data_file_name=None, lightweight_mode=None, py=None, gpu=None, cwd=None, fluent_path=None, topy=None, start_watchdog=None, scheduler_options=None, file_transfer_service=None)#

Bases: object

Instantiates Fluent session within a Slurm environment.

Methods:

__init__([mode, ui_mode, graphics_driver, ...])

Launch Fluent session in standalone mode.

__init__(mode=None, ui_mode=None, graphics_driver=None, product_version=None, dimension=None, precision=None, processor_count=None, journal_file_names=None, start_timeout=-1, additional_arguments='', env=None, cleanup_on_exit=True, start_transcript=True, case_file_name=None, case_data_file_name=None, lightweight_mode=None, py=None, gpu=None, cwd=None, fluent_path=None, topy=None, start_watchdog=None, scheduler_options=None, file_transfer_service=None)#

Launch Fluent session in standalone mode.

Parameters:

modeFluentMode

Launch mode of Fluent to point to a specific session type.

ui_modeUIMode

Fluent user interface mode. Options are the values of the UIMode enum.

graphics_driverFluentWindowsGraphicsDriver or FluentLinuxGraphicsDriver

Graphics driver of Fluent. Options are the values of the FluentWindowsGraphicsDriver enum in Windows or the values of the FluentLinuxGraphicsDriver enum in Linux.

product_versionFluentVersion or str or float or int, optional

Version of Ansys Fluent to launch. To use Fluent version 2024 R2, pass FluentVersion.v242, "24.2.0", "24.2", 24.2, or 242. The default is None, in which case the newest installed version is used.

dimensionDimension or int, optional

Geometric dimensionality of the Fluent simulation. The default is None, in which case Dimension.THREE is used. Options are either the values of the Dimension enum (Dimension.TWO or Dimension.THREE) or any of 2 and 3.

precisionPrecision or str, optional

Floating point precision. The default is None, in which case Precision.DOUBLE is used. Options are either the values of the Precision enum (Precision.SINGLE or Precision.DOUBLE) or any of "double" and "single".

processor_countint, optional

Number of processors. The default is None, in which case 1 processor is used. In job scheduler environments the total number of allocated cores is clamped to value of processor_count.

journal_file_namesstr or list of str, optional

The string path to a Fluent journal file or a list of such paths. Fluent executes the journals. The default is None.

start_timeoutint, optional

Maximum allowable time in seconds for connecting to the Fluent server. The default is 60.

additional_argumentsstr, optional

Additional arguments to send to Fluent as a string in the same format they are normally passed to Fluent on the command line.

envdict[str, str], optional

Mapping to modify environment variables in Fluent. The default is None.

cleanup_on_exitbool, optional

Whether to shut down the connected Fluent session when PyFluent is exited, or the exit() method is called on the session instance, or if the session instance becomes unreferenced. The default is True.

start_transcriptbool, optional

Whether to start streaming the Fluent transcript in the client. The default is True. You can stop and start the streaming of the Fluent transcript subsequently via the method calls, transcript.start() and transcript.stop() on the session object.

case_file_namestr, optional

Name of the case file to read into the Fluent session. The default is None.

case_data_file_namestr, optional

Name of the case data file. If names of both a case file and case data file are provided, they are read into the Fluent session.

lightweight_modebool, optional

Whether to run in lightweight mode. In lightweight mode, the lightweight settings are read into the current Fluent solver session. The mesh is read into a background Fluent solver session, which replaces the current Fluent solver session once the mesh is read and the lightweight settings made by the user in the current Fluent solver session have been applied in the background Fluent solver session. This is all orchestrated by PyFluent and requires no special usage. This parameter is used only when case_file_name is provided. The default is False.

pybool, optional

If True, Fluent will run in Python mode. Default is None.

gpubool, optional

If True, Fluent will start with GPU Solver.

cwdstr, Optional

Working directory for the Fluent client.

fluent_path: str, Optional

User provided Fluent installation path.

topybool or str, optional

A boolean flag to write the equivalent Python journal(s) from the journal(s) passed. Can optionally take the file name of the new python journal file.

start_watchdogbool, optional

When cleanup_on_exit is True, start_watchdog defaults to True, which means an independent watchdog process is run to ensure that any local GUI-less Fluent sessions started by PyFluent are properly closed (or killed if frozen) when the current Python process ends.

scheduler_optionsdict, optional

Dictionary containing scheduler options. Default is None.

Currently only the Slurm scheduler is supported. The scheduler_options dictionary must be of the form {"scheduler": "slurm", "scheduler_headnode": "<headnode>", "scheduler_queue": "<queue>", "scheduler_account": "<account>"}. The keys scheduler_headnode, scheduler_queue and scheduler_account are optional and should be specified in a similar manner to Fluent’s scheduler options.

file_transfer_serviceoptional

File transfer service for uploading and downloading files to and from the server.

Returns:

Union [Meshing, PureMeshing, Solver, SolverIcing, dict]: Session object or configuration dictionary if dry_run = True.

Notes

Job scheduler environments such as SLURM, LSF, and PBS, allocate resources and compute nodes. The allocated machines and core counts are queried from the scheduler environment and passed to Fluent.