Link Search Menu Expand Document

Running mcmicro on the O2 compute cluster

O2 is a high-performance cluster at Harvard Medical School (HMS). Non-HMS users can safely ignore this page.

To run mcmicro on O2, your account must be in the lsp group to have proper permissions to run the module containers. There are several other important caveats.

  1. Installation only requires Nextflow. Docker is not needed, because O2 uses Singularity to execute module containers.

  2. If your account is not in the lsp group (type groups to check), then please run the following command to prepare your environment: nextflow run labsyspharm/mcmicro/setup/

  3. When working with exemplars, please download your own copy to /n/scratch3/users/.../$USER/ (where $USER is your eCommons ID and ... is its first letter).

O2 execution

To run the pipeline on O2, four additional steps are required:

  1. You must load the necessary O2 modules.
  2. All pipeline calls need to have the flag -profile O2. Use -profile O2WSI and -profile O2TMA for very large whole-slide images and tissue microarrays, respectively. The profiles differ in the amount of resources requested for each module.
  3. If your account is not in the lsp group, please use -profile O2ext.
  4. To avoid running over on your disk quota, it is also recommended to use /n/scratch3 for holding the work/ directory. Furthermore, n/scratch3 is faster than /home or /n/groups, so jobs will complete faster.
  5. Instruct Nextflow to generate an execution report in a central location by adding -with-report "/n/groups/lsp/mcmicro/reports/$USER-$(date -Is).html" to your Nextflow command. This will help the staff evaluate resource usage and success/failure rates .

Compose an sbatch script that encapsulates resource requests, module loading and the nextflow command into a single entity. Create a file based on the following template:

#SBATCH -p short
#SBATCH -J nextflow_O2              
#SBATCH -o run.o
#SBATCH -e run.e
#SBATCH -t 0-12:00
#SBATCH --mem=8G
#SBATCH --mail-type=END         # Type of email notification- BEGIN,END,FAIL,ALL
#SBATCH   # Email to which notifications will be sent

module purge
module load java
/home/$USER/bin/nextflow run labsyspharm/mcmicro --in /n/scratch3/users/${USER:0:1}/$USER/exemplar-001 -profile O2 -w /n/scratch3/users/${USER:0:1}/$USER/work

replacing relevant fields (e.g., with your own values.

The pipeline run can then be kicked off with sbatch

Requesting O2 resources

The default profiles O2, O2WSI and O2TMA establish reasonable defaults for O2 resource requests. These will work for most scenarios, but individual projects may also have custom time and memory requirements. To overwrite the defaults, compose a new config file (e.g., myproject.config) specifying the desired custom requirements using process selectors. For example, to request 128GB of memory and 96 hours in the medium queue for ASHLAR, one would specify

  withName:ashlar {
    queue  = 'medium'
    time   = '96h'
    memory = '128G'

Use existing profiles as examples. Once myproject.config is composed, it can be provided to a nextflow run command using the -c flag:

nextflow run labsyspharm/mcmicro --in /path/to/exemplar-001 -profile O2 -c myproject.config

Note that -profile is still needed because it defines additional configurations, such as where to find the container modules. myproject.config simply overrides a portion of the fields in the overall profile.