UPPMAX Configuration

All nf-core pipelines have been successfully configured for use on the Swedish UPPMAX clusters.

Getting help

We have a Slack channel dedicated to UPPMAX users on the nf-core Slack: https://nfcore.slack.com/channels/uppmax

Using the UPPMAX config profile

The recommended way to activate Nextflow, nf-core, and any pipeline available in nf-core on UPPMAX is to use the module system:

# Log in to the desired cluster
ssh <USER>@{rackham,miarka,bianca}.uppmax.uu.se
# Activate the modules, you can also choose to use a specific version with e.g. `Nextflow/21.10`.
module load bioinfo-tools Nextflow nf-core nf-core-pipelines

To use, run the pipeline with -profile uppmax (one hyphen). This will download and launch the uppmax.config which has been pre-configured with a setup suitable for the UPPMAX servers. It will enable Nextflow to manage the pipeline jobs via the Slurm job scheduler. Using this profile, Docker image(s) containing required software(s) will be downloaded, and converted to Singularity image(s) if needed before execution of the pipeline.

Recent version of Nextflow also support the environment variable NXF_SINGULARITY_CACHEDIR which can be used to supply images. Images for some nf-core pipelines are available under /sw/data/ToolBox/nf-core/ and those can be used by NXF_SINGULARITY_CACHEDIR=/sw/data/ToolBox/nf-core/; export NXF_SINGULARITY_CACHEDIR.

In addition to this config profile, you will also need to specify an UPPMAX project id. You can do this with the --project flag (two hyphens) when launching Nextflow. For example:

# Launch a nf-core pipeline with the uppmax profile for the project id snic2018-1-234
$ nextflow run nf-core/<PIPELINE> -profile uppmax --project snic2018-1-234 [...]

NB: If you’re not sure what your UPPMAX project ID is, try running groups or checking SUPR.

Just run Nextflow on a login node and it will handle everything else.

Remember to use -bg to launch Nextflow in the background, so that the pipeline doesn’t exit if you leave your terminal session. Alternatively, you can also launch Nextflow in a screen or a tmux session.

Using AWS iGenomes references

A local copy of the AWS iGenomes resource has been made available on all UPPMAX clusters so you should be able to run the pipeline against any reference available in the conf/igenomes.config. You can do this by simply using the --genome <GENOME_ID> parameter.

Getting more memory

If your nf-core pipeline run is running out of memory, you can run on a fat node with more memory using the following Nextflow flags:

--clusterOptions "-C mem256GB -p node" --max_memory "256GB"

This raises the ceiling of available memory from the default of 128.GB to 256.GB. rackham has nodes with 128GB, 256GB and 1TB memory available.

Note that each job will still start with the same request as normal, but restarted attempts with larger requests will be able to request greater amounts of memory.

All jobs will be submitted to fat nodes using this method, so it’s only for use in extreme circumstances.

Different UPPMAX clusters

The UPPMAX nf-core configuration profile uses the hostname of the active environment to automatically apply the following resource limits:

  • rackham
    • cpus available: 20 cpus
    • memory available: 125 GB
  • bianca
    • cpus available: 16 cpus
    • memory available: 109 GB
  • miarka
    • cpus available: 48 cpus
    • memory available: 357 GB

Development config

If doing pipeline development work on UPPMAX, the devel profile allows for faster testing.

Applied after main UPPMAX config, it overwrites certain parts of the config and submits jobs to the devcore queue, which has much faster queue times.

All jobs are limited to 1 hour to be eligible for this queue and only one job allowed at a time. It is not suitable for use with real data.

To use it, submit with -profile uppmax,devel.

Running on bianca

⚠️ For more information, please follow the following guides:

For security reasons, there is no internet access on bianca so you can’t download from or upload files to the cluster directly. Before running a nf-core pipeline on bianca you will first have to download the pipeline and singularity images needed elsewhere and transfer them via the wharf area to your own bianca project.

In this guide, we use rackham to download and transfer files to the wharf area, but it can also be done on your own computer. If you use rackham to download the pipeline and the singularity containers, we recommend using an interactive session (cf interactive guide), which is what we do in the following guide.

It is recommended to activate Nextflow, nf-core and your nf-core pipeline through the module system (see Using the UPPMAX config profile above). In case you need a specific version of any of these tools you can follow the guide below.

Download and install Nextflow

# Connect to rackham
$ ssh -X <USER>@rackham.uppmax.uu.se
# Or stay in your terminal
# Download the nextflow-all bundle
$ wget https://github.com/nextflow-io/nextflow/releases/download/v<NEXTFLOW_VERSION>/nextflow-<NEXTFLOW_VERSION>-all
# Connect to the wharf area using sftp
$ sftp <USER>-<BIANCA_PROJECT>@bianca-sftp.uppmax.uu.se:<USER>-<BIANCA_PROJECT>
# Transfer nextflow to the wharf area
sftp> put nextflow-<NEXTFLOW_VERSION>-all .
# Exit sftp
$ exit
# Connect to bianca
$ ssh -A <USER>-<BIANCA_PROJECT>@bianca.uppmax.uu.se
# Go to your project
$ cd /castor/project/proj_nobackup
# Make folder for Nextflow
$ mkdir tools
$ mkdir tools/nextflow
# Move Nextflow from the wharf area to its directory
$ mv /castor/project/proj_nobackup/wharf/<USER>/<USER>-<BIANCA_PROJECT>/nextflow-<NEXTFLOW_VERSION>-all /castor/project/proj_nobackup/tools/nextflow
# Establish permission
$ chmod a+x /castor/project/proj_nobackup/tools/nextflow/nextflow-<NEXTFLOW_VERSION>-all
# If you want other people to use it
# Be sure that your group has rights to the directory as well
$ chown -R .<BIANCA_PROJECT> /castor/project/proj_nobackup/tools/nextflow/nextflow-<NEXTFLOW_VERSION>-all
# Make a link to it
$ ln -s /castor/project/proj_nobackup/tools/nextflow/nextflow-<NEXTFLOW_VERSION>-all /castor/project/proj_nobackup/tools/nextflow/nextflow
# And every time you're launching Nextflow, don't forget to export the following ENV variables
# Or add them to your .bashrc file
$ export NXF_HOME=/castor/project/proj/nobackup/tools/nextflow/
$ export PATH=${NXF_HOME}:${PATH}
$ export NXF_SINGULARITY_CACHEDIR=/castor/project/proj_nobackup/singularity-images

Install nf-core tools

# Connect to rackham
$ ssh -X <USER>@rackham.uppmax.uu.se
# Or stay in your terminal
# Install the latest pip version
$ pip3 install --upgrade --force-reinstall git+https://github.com/nf-core/tools.git@dev --user

Download and transfer a nf-core pipeline

# Connect to rackham
$ ssh -X <USER>@rackham.uppmax.uu.se
# Or stay in your terminal
# Open an interactive session (if you are on rackham)
$ interactive <rackham_project>
# Download a pipeline with the singularity images
$ nf-core download <PIPELINE> -r <PIPELINE_VERSION> -s --compress none
# If necessary, extra singularity images can be download separately
# For example, if you downloaded nf-core/sarek, you will need extra images for annotation
# Here we download the nf-core/sarek GRCh38 specific images
$ singularity pull --name nfcore-sareksnpeff-2.7.GRCh38.img docker://nfcore/sareksnpeff:2.7.GRCh38
$ singularity pull --name nfcore-sarekvep-2.7.GRCh38.img docker://nfcore/sarekvep:2.7.GRCh38
# Which can then be moved into the nf-core/sarek download folder
$ mv *.img nf-core-sarek-2.7/singularity-images/.
# Connect to the wharf area using sftp
$ sftp <USER>-<BIANCA_PROJECT>@bianca-sftp.uppmax.uu.se:<USER>-<BIANCA_PROJECT>
# Transfer <PIPELINE> folder from rackham to the wharf area
sftp> put -r nf-core-<PIPELINE>-<PIPELINE_VERSION> .
# The archives will be in the wharf folder in your user home on your bianca project
# Connect to bianca
$ ssh -A <USER>-<BIANCA_PROJECT>@bianca.uppmax.uu.se
# Go to your project
$ cd /castor/project/proj_nobackup
# Make and go into a nf-core directory (where you will store all nf-core pipelines')
$ mkdir nf-core
$ cd nf-core
# Move the folder from the wharf area to the project
$ cp /castor/project/proj_nobackup/wharf/<USER>/<USER>-<BIANCA_PROJECT>/nf-core-<PIPELINE>-<PIPELINE_VERSION> .
# If you want other people to use it,
# Be sure that your group has rights to the directory as well
# Make a symbolic link to the extracted repository
$ ln -s nf-core-<PIPELINE>-<PIPELINE_VERSION> nf-core-<PIPELINE>-default

The principle is to have every member of your project to be able to use the same nf-core/<PIPELINE> version at the same time. So every member of the project who wants to use nf-core/<PIPELINE> will need to do:

# Connect to bianca
$ ssh -A <USER>-<BIANCA_PROJECT>@bianca.uppmax.uu.se
# Go to your user directory
$ cd /home/<USER>
# Make a symbolic link to the default nf-core/<PIPELINE>
$ ln -s /castor/project/proj_nobackup/nf-core/nf-core-<PIPELINE>-default nf-core-<PIPELINE>

And then nf-core/<PIPELINE> can be used with:

# run <PIPELINE> on bianca
$ nextflow run ~/<PIPELINE> -profile uppmax --project <BIANCA_PROJECT> --genome <GENOME_ASSEMBLY> ...

Update a pipeline

To update, repeat the same steps as for installing and update the link.

# Connect to bianca (Connect to rackham first if needed)
$ ssh -A <USER>-<BIANCA_PROJECT>@bianca.uppmax.uu.se
# Go to the nf-core directory in your project
$ cd /castor/project/proj_nobackup/nf-core
# Remove link
$ unlink nf-core-<PIPELINE>-default
# Link to new nf-core/<PIPELINE> version
$ ln -s nf-core-<PIPELINE>-<PIPELINE_VERSION> nf-core-<PIPELINE>-default

You can for example keep a nf-core-<PIPELINE>-default version that you are sure is working, an make a link for a nf-core-<PIPELINE>-testing or nf-core-<PIPELINE>-development.

Config file

See config file on GitHub

// UPPMAX Config Profile
params {
    // Description is overwritten for other clusters below
    config_profile_description = 'UPPMAX (Bianca) cluster profile provided by nf-core/configs.'
    config_profile_contact = 'Phil Ewels (@ewels)'
    config_profile_url = 'https://www.uppmax.uu.se/'
    project = null
    clusterOptions = null
    schema_ignore_params = "genomes,input_paths,cluster-options,clusterOptions,project"
    validationSchemaIgnoreParams = "genomes,input_paths,cluster-options,clusterOptions,project,schema_ignore_params"
    save_reference = true
    // Defaults set for Bianca - other clusters set below
    max_memory = 500.GB
    max_cpus = 16
    max_time = 240.h
    // illumina iGenomes reference file paths on UPPMAX
    igenomes_base = '/sw/data/igenomes/'
singularity {
    enabled = true
    envWhitelist = 'SNIC_TMP'
def hostname = "r1"
try {
    hostname = ['/bin/bash', '-c', 'sinfo --local -N -h | grep -m 1 -F -v CLUSTER: | cut -f1 -d" "'].execute().text.trim()
} catch (java.io.IOException e) {
    System.err.println("WARNING: Could not run sinfo to determine current cluster, defaulting to rackham")
// closure to create a suitable clusterOptions
def clusterOptionsCreator = { m ->
    String base = "-A $params.project ${params.clusterOptions ?: ''}"
    // Do not use -p node on irma or if a thin node/core is enough
    if (m <= 125.GB || hostname.startsWith("i")) {
        return base
    // cluster is miarka
    if (hostname.startsWith("m")) {
        // job will fit on a regular node
        if (m <= 357.GB) {
            return base
        // job requires at least a 2TB node
        if (m <= 2000.GB) {
            return base + " --mem 2TB "
        // job requires the largest node
        return base + " -C mem4TB "
    if (m <= 250.GB) {
        return base + " -p node -C mem256GB "
    // Use mem1TB for remaining cases on rackham (no 512 Gbyte nodes)
    if (hostname.startsWith("r")) {
        return base + " -p node -C mem1TB "
    if (m > 500.GB) {
        // Special case for snowy very fat node (only remaining case that's above 500 GB)
        return base + " -p veryfat "
    // Should only be cases for mem512GB left (snowy and bianca)
    return base + " -p node -C mem512GB "
process {
    executor = 'slurm'
    clusterOptions = { clusterOptionsCreator(task.memory) }
    // Use node local storage for execution.
    scratch = '$SNIC_TMP'
// Cluster: Snowy
//   Caution: Bianca nodes will be project name-nodenumber, e.g. sens2021500-001
//   so cannot rely on just starting with 's'
if (hostname.matches("^s[0-9][0-9]*")) {
    params.max_time = 700.h
    params.max_memory = 3880.GB
    params.config_profile_description = 'UPPMAX (Snowy) cluster profile provided by nf-core/configs.'
// Cluster: Irma
if (hostname.startsWith("i")) {
    params.max_memory = 250.GB
    params.config_profile_description = 'UPPMAX (Irma) cluster profile provided by nf-core/configs.'
// Cluster: Miarka
if (hostname.startsWith("m")) {
    params.max_memory = 357.GB
    params.max_cpus = 48
    params.max_time = 480.h
    params.config_profile_description = 'UPPMAX (Miarka) cluster profile provided by nf-core/configs.'
// Cluster: Rackham
if (hostname.startsWith("r")) {
    params.max_cpus = 20
    params.max_memory = 970.GB
    params.config_profile_description = 'UPPMAX (Rackham) cluster profile provided by nf-core/configs.'
// Cluster: Bianca - set in initial params block above
// Additional devel profile for running in devel queue
// Run with `-profile upppmax,devel`
profiles {
    devel {
        params {
            config_profile_description = 'Testing & development profile for UPPMAX, provided by nf-core/configs.'
            // Max resources to be requested by a devel job
            max_memory = 120.GB
            max_time = 1.h
        executor.queueSize = 1
        process.queue = 'devel'