Workflow Run Script

Hutch Data Core

Updated: April 23, 2021

Edit this Page via GitHub       Comment by Filing an Issue      Have Questions? Ask them here.

Each workflow which you run with Nextflow should ideally be run from its own directory. The reason we recommend this is that Nextflow uses a small database in the working directory to keep track of what tasks it has run. This is particularly useful if your workflow is interrupted and you want to re-run it while picking up from where you left off using the resume function.

To keep your workflows organized, we recommend creating a folder for each run of a workflow, and placing in that folder a “run script” which contains the nextflow run ... command, along with all of the flags and parameters you use. This makes it easy to come back to a workflow and know exactly what you ran. It also allows you to re-run that workflow knowing that the -resume function will work as intended.

Here is an example of a run script, with more details and explanation below:

#!/bin/bash

set -Eeuo pipefail

# Nextflow Version
NXF_VER=20.10.0

# Nextflow Configuration File
NXF_CONFIG=/path/to/your/nextflow.config

# Workflow to Run (e.g. GitHub repository)
WORKFLOW_REPO=FredHutch/workflow-repository

# Queue to use for exeution
QUEUE=campus-new

# Workflow-Specific Options
INPUT_FOLDER=/path/to/input/folder/for/this/analysis/
OUTPUT_FOLDER=/path/to/output/folder/for/this/analysis/

# Load the Nextflow module (if running on rhino/gizmo)
ml Nextflow

# Load the Singularity module (if running on rhino/gizmo with Singularity)
ml Singularity
# Make sure that the singularity executables are in the PATH
export PATH=$SINGULARITYROOT/bin/:$PATH

# Run the workflow
NXF_VER=$NXF_VER \
nextflow \
    run \
    -c ${NXF_CONFIG} \
    ${WORKFLOW_REPO} \
    --input_folder $INPUT_FOLDER \
    --output_folder $OUTPUT_FOLDER \
    -process.queue $QUEUE \
    -with-report nextflow.report.html \
    -resume

Header: set directive

The line set -Eeuo pipefail is used to ensure that any errors that are encountered while executing the BASH script will cause the entire process to halt. This is particularly useful if you happen to have a typo (e.g. INPUT_FOLDER vs. INPT_FOLDER) which would otherwise result in an empty string being unintentionally passed in as an argument to run the workflow.

Nextflow Version

By setting NXF_VER, we control which version of Nextflow is used to execute the workflow. By placing this directive in the run script we can ensure that future attempts to re-run the same workflow will also use the same version of Nextflow for an added level of reproducibility.

Nextflow Configuration File

Use NXF_CONFIG to specify the configuration file used to run workflows either on SLURM/gizmo or on AWS.

Workflow Repository

The WORKFLOW_REPO variable indicates the GitHub repository being used to host the workflow. For additional control over a particular release or branch to execute within that repository, use the -r flag.

Execution Queue

The QUEUE variable can be used to specify which compute queue to use for executing jobs. On gizmo, this is campus-new, while on AWS Batch the name of the queue may depend on the setup of your particular lab’s account.

Workflow Specific Options

Each workflow requires a different set of options or parameters.

NOTE: Parameters which are specific to a workflow have two dashes (--), while parameters which modify the behavior of Nextflow itself have a single dash (-).

We recommend setting up all workflow parameters in a block of environment variables as shown in this run script, as it makes it more readable when modifying the script in the future.

Load the Nextflow Module

When running on a rhino or gizmo node, the Nextflow codebase can be loaded with ml Nextflow. If you are running locally this can be omitted, but you will also need to install Nextflow on that local machine

Load the Singularity Module

When running on gizmo and using Singularity to execute within containers, you need to load Singularity on the head node with ml Singularity. If either of those are not the case (not running on gizmo or not using Singularity), this can be omitted.

NOTE: You should not run any workflows which use Singularity from any rhino node.

Nextflow Run - Formatting

The run script template is set up in such a way that you will not have to modify the nextflow run command itself, instead just modifying the environment variables which are then populated into the command.

However, when adapting this template for your own workflow you will necessarily have to format additional parameters which are specific to your workflow. In other words, --input_folder and --output_folder are only given as an example, and you will have to change those lines (and add more) for your own workflow.

When modifying the run script, the single biggest source of errors has to do with the backslashes used to escape the line endings. To break this down, everything in the template script above between NXF_VER and -resume is actually on a single line. While the command is written on multiple lines, the backslash (\) character is used to escape that line ending and have the BASH interpreter treat those all as though they were on a single line.

The short summary is: when you modify the script, make sure to include a backslash for any new lines that you add, and make doubly sure that there are no characters (including spaces!) to the right of that backslash.

Execution Report

The -with-report flag creates a simple HTML page which summarizes the execution of the workflow. If you re-run a workflow, the previous report will be renamed "*.html.1", "*.html.2", etc. This file is a very helpful summary of the complete set of parameters, data, and code used to run a workflow.

Other Useful Flags

Limit the number of jobs with -process.maxForks

The flag -process.maxForks accepts an integer which specifies the maximum number of jobs which will be submitted for execution concurrently per process. This can be very helpful if you want to limit your use of a shared queue.

Complete Documentation

For a complete list of the options available to invoke Nextflow, please read their online documentation.

Updated: April 23, 2021

Edit this Page via GitHub       Comment by Filing an Issue      Have Questions? Ask them here.