CSCS

The following section provides details on how to run Sencast on the supercomputer Piz Daint at CSCS.

Register for an account

Get access permission to Daint from your local IT Admin. You will be required to set up multifactor authentication

Access using Jupyter

Access using ssh

This access is only valid for 24 hours after which the process will need to be repeated. For details on how to automate see here: https://user.cscs.ch/access/auth/mfa and for Windows see here: https://user.cscs.ch/access/auth/mfa/windows

  • Login at https://sshservice.cscs.ch/

  • Click Get a signed key follow the instructions and download the private and public key

  • Move the keys to your users .ssh directory

mv /downloads/location/cscs-key-cert.pub ~/.ssh/cscs-key-cert.pub
mv /download/location/cscs-key ~/.ssh/cscs-key
chmod 0600 ~/.ssh/cscs-key

  • Login to the CSCS entrance server

ssh -A username@ela.cscs.ch

  • Switch to Piz Daint

ssh username@daint.cscs.ch

Install Sencast

This step must be completed on the command line after logging into Piz Daint using one of the above methods.

Load the required modules

module load daint-mc
module load sarus

Clone the repo for sencast to your user area:

cd ~
git clone https://github.com/eawag-surface-waters-research/sencast.git

Update the environment and parameters scripts that you want to run.

Pull the image you want from dockerhub:

srun -C mc -A em09 sarus pull --login eawag/sencast:0.0.2

then enter your credentials for the repository (There is no prompt)

<username> <password>

The docker image (now for sarus) is automatically saved in ${SCRATCH}/.sarus

If this fails try running the command again.

Run Sencast

Move to the scratch drive and create an output folder don’t save large amounts of data to user area Data stored in the scratch drive is removed after 30 days.

cd ${SCRATCH}
mkdir DIAS

Create a submission script containing the following (adjust details to match your user) - make sure you are writing to scratch.

vim run.sh

#!/bin/bash -l
#SBATCH --job-name="sencast"
#SBATCH --account="em09"
#SBATCH --mail-type=ALL
#SBATCH --mail-user=username@eawag.ch
#SBATCH --time=24:00:00
#SBATCH --nodes=1
#SBATCH --ntasks-per-core=1
#SBATCH --ntasks-per-node=1
#SBATCH --cpus-per-task=36
#SBATCH --partition=normal
#SBATCH --constraint=mc
#SBATCH --hint=nomultithread
export OMP_NUM_THREADS=$SLURM_CPUS_PER_TASK
module load daint-mc
module load sarus
image='eawag/sencast:0.0.2'
envvars='docker.ini'
params='parameters.ini'
filepath="${SCRATCH}/DIAS"

cd ~/sencast
srun sarus run --mount=type=bind,source=${filepath},destination=/DIAS --mount=type=bind,source=$(pwd),dst=/sencast ${image} -e ${envvars} -p ${params}

:w save file

:q exit vim

Then you can run Sencast:

sbatch run.sh

See the status of your job:

squeue -u username

You get an email when the job begins and if it fails. A live log is deposited in the directory from where you start the run.