Resources for testing

Test runners on the online nf-neuro repository restrain their resources usage to specifically what is asked for by Nextflow through the configuration files. For nf-neuro tests, you can find this configuration in tests/config/nextflow.config.

If any of your tests ask for more, you need to define it correctly.

Note

Usually, containers detects out-of-bound resource consumption automatically. When using the devcontainer or configuring Nextflow to run using docker or apptainer, tests will fail if they haven’t been assigned them sufficiently. It can be a good way to see if your workload needs more resources quickly !

Available runner classes

Runner Class	CPUs	RAM
scilus-docker (default)	4	4G
scilus-docker-large	6	32G

Request for more test resources

Caution

Specialized runners are limited !!! They are allocated for hungry processes, such as AI/ML models and large dataset studies, don’t abuse them. The more they are requested for, the longer PR take to merge. Don’t ask for them for nothing, meaning be smart in designing your modules ! PRs deemed not needing them will be automatically closed.

Custom resource requirements need to be defined in nextflow.config file(s) associated to your tests and/or test cases. In them, add the process.cpu and process.memory and set their requirements as needed :

process.memory = '9G'
process.cpus = 6

Select an appropriate runner class for your tests. To do so, edit github/workflows/run_checks_suite.yml. Find the matrix definition for the nf-test job (currently around line 184) and add the following request for a runner in the include section :

- runner: <name-of-the-runner-class>
  path: modules/nf-neuro/<category>/<tool>

You should not need more resources for subworkflows !

Subworkflows use stub-runs to test their integration, but they don’t run the actual workload of their modules. They thus should not ask for more resources than what’s needed to call help commands and create empty files.

There are definitely use-cases, so if you think otherwise, before implementing the resource request above, open a discussion with the maintainers !