![]() ![]() These paths can be relative or absolute relative paths are considered relative to the DIVCFG file. The template file provides a skeleton that divvy will populate with job-specific attributes. The submission_template attributeĮach compute package specifies a path to a template file ( submission_template). We are telling divvy that submitting this job should be done with: sbatch submission_script.txt. For example, in our compute package named develop_package, we've set submission_command to sbatch. The submission_command attribute is the string your cluster resource manager uses to submit a job. There are only two required attributes: each compute package must specify the submission_command and submission_template attributes. You can also add whatever attributes you like to the compute package. You can make as many compute packages as you wish, and name them whatever you wish. You can activate any other compute package on the fly by calling the activate_package function. You can make your default whatever you like. If you don't specify a package to activate, divvy uses the package named default. Divvy uses these compute packages to determine how to submit your jobs. The sub-sections below compute_packages each define a compute package that can be activated. Submission_template: templates/slurm_template.sub Submission_template: templates/local_template.sub Here is an example divvy configuration file: compute_packages: The DIVCFG file is a yaml file listing different compute packages. If it cannot find one there, then it will load a default configuration file with a few basic compute packages. If you don't specify one, divvy will first look for a file in the $DIVCFG environment variable. If using divvy from within python, you can pass a configuration file when you construct a new ComputingConfiguration object. The file specifying the compute packages is called the DIVCFG file. You can customize your compute packages as much as you need. By just choosing local or slurm you can change where your job is run. Each package represents a computing resource for example, by default we have 1 package (called local) that populates templates to simple run jobs in the local console, and another package (called slurm) with a generic template to submit jobs to a SLURM cluster resource manager. ![]() Divvy configuration files The DIVCFG environment variableĪt the heart of divvy is a yaml configuration file that specifies your available compute_packages. ![]()
0 Comments
Leave a Reply. |
AuthorWrite something about yourself. No need to be fancy, just an overview. ArchivesCategories |