job-farmer 0.1.8

A simple job farmer for HPC platforms

Job Farmer

Description

Job Farmer is a python API designed to prepare jobs for HPC platforms supporting SLURM. It collects the job configurations of the associated tasks and prepares the task list and slurm configuration files based on the system constraints and the required resources.

Installation

From PyPI

pip install job-farmer

From sources

git clone [e-mail]:riffard/job_farmer.git
cd job_farmer
pip install .

Usage

Job configuration

In the file test/simple_job.py, we are showing a simple job submission.

First, you import the job-farmer:

import job_farmer

Then you create the System object (NERSC/CORI-Haswell in this example) and the Farmer object

system = job_farmer.CoriHaswell()

farmer = job_farmer.Farmer(system)

The nex step consists on creation the JobConfiguration and add it to the Farmer:

job_configuration = job_farmer.JobConfiguration(name='test_job',
                       account='your_account',
                       queue='queue',
                       duration='xx:xx:xx')

farmer.add_global_job_configuration(job_configuration)

You add the Tasks to the Farmer and you link them to the job_configuration

task = job_farmer.Task(commands=[f'export MY_MESSAGE="Hello World!"',
                                 'echo "${MY_MESSAGE}"'], 
                        n_threads=1)
farmer.add_task(job_configuration.name, task)

And it builds the task list and the slurm configuration file:

farmer.build()

Finally, you can submit the jobs to the batch system using:

farmer.submit(dry_run = False)

(by default, dry_run = True).

Note: By default, the job-farmer logs are not displayed. To display them, you need to import the logging package and set the log level to logging.INFO such as:**

logging.basicConfig(level=logging.INFO)

Running jobs

Job runner

Batch submission

Support

If you are encountering any issue, please open a git issue: here

Contributing

Any contribution is more than welcome. To add any new feature, please follow this procedure:

1. Create a git issue here
2. Create a new branch with the following format: [initials]_[feature title]_issue_[issue number].
3. Once the development is nearly completed, push the branch to the repository and create a merge request with the following guideline:
   1. The MR title must contain a short feature/bugfix description
   2. The MR description must contain a detailed description of the feature/bug fix. Add here any relevant resources.
4. After reviewing the merge request, it will be merged to the main branch.

Adding a new data center and a new system

1. To add a new data center, create a new python file in job_farmer/system with the name of the data center in the filename (use lower case).
2. To add a new system, add a new System class that inherits from System in the correct data center file. To see an example, take a look at job_farmer/system/nersc.py

License

This software is licenced under: GNU Affero General Public License v3.0