Episode 41: New syntax in Nextflow 24.04.
Join us in exploring the latest Nextflow release, 24.04. Phil Ewels and Ben Sherman dive deep into the new updates in Nextflow’s most recent version. We discuss advanced retry strategies, job arrays, resourceLimits, Singularity’s OCI mode, and the game-changing Workflow Output Definition.
We’re actively seeking feedback from the community on these new features. We would love for you to experiment with the new syntax and let us know your thoughts.
Please let us know what you think either via Nextflow GitHub issues or via the community forum.
In this episode we refer to a the recent blog post about the release, which you can find here: Nextflow 24.04 - Release highlights.
Intro - tweaks, fixes and new support in 24.04
Performance and Stability Improvements
Ben kicks off the discussion with key performance and stability improvements. The latest stable release, 24.04, includes numerous bug fixes and performance enhancements. There’s a significant focus on closing gaps and adding retry strategies for better stability, especially concerning cloud providers. Notably, if Nextflow’s API calls against cloud providers fail for any server-side reason, it will retry automatically, avoiding pipeline failures caused by temporary server issues.
Publishing Changes
Previously, if a file failed to publish, Nextflow would only issue a warning. It was possible for a pipeline to complete successfully without noticing missing output files. Now, by default, Nextflow will fail the pipeline if the publishing fails, though there’s an option to revert to the old behavior. Additionally, retry strategies for publishing have been introduced to ensure retry attempts if an issue arises.
Job Arrays
One highly requested feature is the introduction of job arrays. This allows users to submit many jobs as a single submission, alleviating strain on schedulers. The submission happens in a batch, and then the scheduler can process and plan effectively. Once a job is submitted as part of a job array, the jobs run independently. If any child job fails, it is resubmitted without affecting the rest of the array.
Singularity OCI mode and GA4GH TES
In the past, Singularity supported Docker images by converting them into SIF files, consuming storage and time. Now, both Singularity and Apptainer can run OCI images directly, saving valuable resources.
Additionally, the TES executor has seen significant improvements, now supporting TES 1.1, which brings broader compatibility and integration with existing workflows.
Major new syntax features
Topic Channels
A new feature called topic channels offers a more straightforward approach to collecting channel outputs across pipelines. Channels can emit data to a named topic, simplifying the collection and use of version information from various processes within Nextflow pipelines.
Eval Outputs
Eval outputs simplify the addition of shell commands to tasks. With eval outputs, necessary post-task commands can be defined neatly, avoiding repetitive code within process definitions.
Workflow Output Definition
The concept of workflow output definitions has been introduced. This new syntax streamlines the publication of files by defining publish targets within workflows. Instead of defining publication behaviors within process definitions, users can now manage them at the workflow level, ensuring better clarity and fewer repetitions in the code.
