Caelus Tasks¶

CPL provides a tasks interface to automate various aspects of the CFD simulation workflow that can be executed by calling caelus tasks (see tasks documentation).

Quick tutorial¶

The tasks interface requires a list of tasks provided in a YAML-formatted file as shown below (download):

tasks:
  - clean_case:
      remove_zero: no
      remove_mesh: yes

  - run_command:
      cmd_name: blockMesh

  - run_command:
      cmd_name: pisoSolver

  - process_logs:
      log_file: pisoSolver.log
      plot_residuals: true
      residuals_plot_file: residuals.pdf
      residuals_fields: [Ux, Uy]

The file lists a set of actions to be executed sequentially by caelus tasks. The tasks can accept various options that can be used to further customize the workflow. A sample interaction is shown below

$ caelus -v tasks -f caelus_tasks.yaml
INFO: Caelus Python Library (CPL) v0.1.0
INFO: Caelus CML version: 7.04
INFO: Loaded tasks from: cavity/caelus_tasks.yaml
INFO: Begin executing tasks in cavity
INFO: Cleaning case directory: cavity
INFO: Executing command: blockMesh
INFO: Executing command: pisoSolver
INFO: Processing log file: pisoSolver.log
INFO: Saved figure: cavity/residuals.pdf
INFO: Residual time history saved to residuals.pdf
INFO: Successfully executed 4 tasks in cavity
INFO: All tasks executed successfully.

For a comprehensive list of task file examples, please consult the run_tutorial.yaml files in the tutorials directory of Caelus CML distribution. In particular, the motorBike case provides an example of a tasks workflow involving two different case directories.

Note

When using CPL with OpenFOAM, please replace pisoSolver with pisoFoam etc. Other than the name of the solvers, the rest of the commands do not change when switching between CML and OpenFOAM.

Tasks reference¶

This section documents the various tasks available currently within CPL and the options that can be used to customize execution of those tasks.

The task file must be in YAML format, and must contain one entry tasks that is a list of tasks to be executed.
The tasks are executed sequentially in the order provided until an error is encountered or all tasks are executed successfully.
The tasks must be invoked from within a valid Caelus case directory (see task_set for an exception). All filenames in the task file are interpreted relative to the execution directory where the command is invoked.

run_command – Run OpenFOAM or CML executables¶

This task type is used to execute an OpenFOAM or Caelus CML executable (e.g., blockMesh or pimpleSolver). CPL will ensure that the appropriate version of OpenFOAM or CML is selected and the runtime enviornment is setup properly prior to executing the task. The task must provide one mandatory parameter run_command.cmd_name that is the name of the OpenFOAM or CML executable. Several other options are available and are documented below. Example:

- run_command:
    cmd_name: potentialSolver
    cmd_args: "-initialiseUBCs -noFunctionObjects"
    parallel: true

run_command.cmd_name¶: The name of the CML executable. This option is mandatory.

run_command.cmd_args¶: Extra arguments that must be passed to the OpenFOAM or CML executable. It is recommended that arguments be enclosed in a double-quoted string. Default value is an empty string.

run_command.log_file¶: The filename where the output of the command is redirected. By default, it is the OpenFOAM or CML executable name with the .log extension appended to it. The user can change this to any valid filename of their choice using this option.

run_command.parallel¶: A Boolean flag indicating whether the executable is to be run in parallel mode. The default value is False. If parallel is True, then the default options for job scheduler are used from CPL configuration file, but can be overriden with additional options to run_command.

run_command.num_ranks¶: The number of MPI ranks for a parallel run.

run_command.mpi_extra_args¶

Extra arguments to be passed to mpiexec command (e.g., hostfile options). As with cmd_args, enclose the options within quotes.

- run_command:
    cmd_name: pimpleSolver
    parallel: true
    mpi_extra_args: "--hostfile my_hostfile"

run_python – Run user python scripts¶

This task will run a python script within the case directory. Like run_command task, CPL will ensure that the apporpriate OpenFOAM or CML version as well as python environment is setup correctly prior to invoking the script. The task must provide one mandatory parameter run_python.script that contains the path to the python script to be executed. User can pass additional options to this task as documented below.

- run_python:
    script: "../mytestscript.py"
    script_args: "-v -f option1 arg1 arg2"
    log_file: "mycustom.log"

run_python.script¶: Path to the custom python script.

run_python.script_args¶: Arguments that must be passed to the python script during invocation. As with run_command quote the arguments appropriately to ensure that it is passed correctly to the script.

run_python.log_file¶: The filename where the outputs and error messages from the script is redirected.

copy_files – Copy files¶

This task copies files in a platform-agnostic manner.

copy_files.src¶: A unix-style file pattern that is used to match the pattern of files to be copied. The path to the files must be relative to the execution directory, but can exist in other directories as long as the relative paths are provided correctly. If the pattern matches multiple files, then copy_files.dest must be a directory.

copy_files.dest¶: The destination where the files are to be copied.

copy_tree – Recursively copy directories¶

This task takes an existing directory (src) and copies it to the destination. Internally, this task uses copytree function to copy the directory, please refer to Python docs for more details.

Warning

If the destination directory already exists, the directory is deleted before copying the contents of the source directory. Currently, this task does not provide a way to copy only non-existent files to the destination. Use with caution.

copy_tree.src¶: The source directory that must be recursively copied.

copy_tree.dest¶: The pathname for the new directory to be created.

copy_tree.ignore_patterns¶: A list of Unix-style file patterns used to ignore files present in source directory when copying it to destination. A good example of this is to prevent copying the contents of polyMesh when copying the contents of constant from one case directory to another.

copy_tree.preserve_symlinks¶: A Boolean flag indicating whether symbolic links are preserved when copying. Linux and Mac OSX only.

clean_case – Clean a case directory¶

Use this task to clean up a case directory after a run. By default, this task will preserve all YAML and python files found in the case directory as well as the 0/ directory. For example,

- clean_case:
    remove_zero: yes
    remove_mesh: no
    preserve: [ "0.org" ]

clean_case.remove_zero¶: Boolean flag indicating whether the 0/ directory should be removed. The default value is no.

clean_case.remove_mesh¶: Boolean flag indicating whether the constant/polyMesh directory should be removed. The default value is no.

clean_case.remove_time_dirs¶: Boolean flag indicating whether time directories from a previous run should be removed. The default value is no.

clean_case.remove_processors¶: Boolean flag indicating whether processor directories from a previous run should be removed. The default value is no.

clean_case.purge_generated¶: A Boolean flag when enabled will set remove_time_dirs and remove_processors to yes.

clean_case.purge_all¶: A shortcut to set remove_zero, remove_mesh, remove_time_dirs and remove_processors to yes.

clean_case.preserve¶: A list of Unix-style file patterns that match files that should be preserved within the case directory.

process_logs – Process solver outputs¶

This task takes one mandatory argument log_file that contains the outputs from a CFD run. The time-histories of the residuals are extracted and output to files that can be loaded by gnuplot, or loaded in python using loadtxt command or using Pandas library. Users can also plot the residuals by using the plot_residuals option. For example,

- process_logs:
  log_file: pimpleSolver.log
  log_directory: pimpleSolver_logs

- process_logs:
  log_file: simpleSolver.log
  plot_residuals: yes
  residuals_plot_file: residuals.pdf
  residuals_fields: [Ux, Uy, p]

process_logs.log_file¶: The filename containing the solver residual ouputs. This parameter is mandatory.

process_logs.logs_directory¶: The directory where the processed residual time-history outputs are stored. Default: logs within the execution directory.

process_logs.plot_residuals¶: Boolean flag indicating whether a plot of the convergence time-history is generated. Default value is False.

process_logs.residuals_plot_file¶: The file where the plot is saved. Default value is residuals.png. The user can use an appropriate extension (e.g., .png, .pdf, .jpg) to change the image format of the plot generated.

process_logs.residual_fields¶: A list of fields that are plotted. If not provided, all fields available are plotted.

process_logs.plot_continuity_errors¶: A Boolean flag indicating whether time-history of continuity errors are plotted along with residuals.

task_set – Group tasks¶

A task_set groups a sub-set of tasks that can be executed in a different case directory. Download an example.

task_set.case_dir¶: The path to a valid case directory where the sub-tasks are to be executed. This parameter is mandatory.

task_set.name¶: A unique name to identify this task group.

task_set.tasks¶: The list of sub-tasks. This list can contain any of the tasks that have been documented above.

Table of Contents

Previous topic

Next topic

This Page