NOTS

• Author: Fu Yin

• Update: Dec 15, 2022

• Reading: 10 min

---

Apply an Account

Set up on the cluster (NOTS) please follow the instructions to apply for an account, https://kb.rice.edu/108221, and put your advisor down as your sponsor if your PI has the account.

Fig. 32 NOTS Account Application

File Management

The following is my data managerment, more info please refer to Rice CRC Tutorial.

Fig. 33 NOTS Storage

Warning

• /scratch is not permanent storage, which is to be used only for job I/O. Delete everything you do not need for another run at the end of the job or move to /storage/hpc/work for analysis. Staff may periodically delete files from the /scratch file system even if files are less than 14 days old.

• /storage/hpc/work is only available on the login nodes. Data can be copied between /storage/hpc/work and /scratch before/after running jobs.

Computing Resources

Fig. 34 NOTS Resources

Teriminal Configuraion

NOTS is a Red Hat linux system, the default shell is bash. Becasue I’m not the administrator, so I don’t have the root permissions, which means I can’t use yum to install packages.

In addition, I can’t change the default shell from bash to zsh by myself, but I let the administrator change it to zsh which is a old version. So I install zsh by myself, and auto-run new zsh when I login.

# use wget to download source package 
wget -O zsh.tar.xz https://sourceforge.net/projects/zsh/files/latest/download

# decompress the package
xz -d zsh.tar.xz
tar -xvf zsh.tar

# configuration
cd zsh
./configure --prefix=/storage/hpc/work/ja62/fy21/software/zsh/bin/zsh

# install
make 
make install

# auto launch when login bash
vim ~/.bashrc
export PATH=/storage/hpc/work/ja62/fy21/software/zsh/bin:$PATH

How to start new zsh when login?

vim ~/.zshrc

# Output:
# [1].run Fu's zsh when login at the first time
ZshName=$(which zsh)
if [ "$ZshName" = "/usr/bin/zsh" ];then
  export PATH=/storage/hpc/work/ja62/fy21/software/zsh/bin:$PATH
  echo "Now run Fu's zsh...\nVersion: $(zsh --version)\nPath: $(which zsh)\n"
  zsh
fi
export PATH=/storage/hpc/work/ja62/fy21/software/zsh/bin:$PATH


# [2].source global definitions
if [ -f /etc/zshrc ]; then
	. /etc/zshrc
fi

Some packages I have installed:

Name	Purpose	Url	Date
miniconda
``
``

Conda list:

scientific computation-related package:

Name	Purpose	Way
numpy	…	conda
matplotlib	…	conda
``
``

Job Example

The following is an example named with jobs.slurm, and some common command to run the job.

# submit a job
sbatch jobs.slurm

# check the state of job
sacct -j JobID

The jobs.slurm content:

#!/bin/bash 
##################################### 
# Mission name
#SBATCH --job-name=fy-test

# Use only single node 
#SBATCH -N 1 

# SMP execution, only one task 
#SBATCH -n 1 

# Use physical cores only not hyperthreads 
#SBATCH --threads-per-core=1 

# Request multiple cores 
#SBATCH --cpus-per-task=1 

# Time for your request 
#SBATCH --time=00:10:10


#SBATCH --gres=gpu:1 


# load package
module load GCC

# run job
srun echo "ssss" >> 2.txt
srun python ./mission.py

Warning

Note there is no blank space between # and SBATCH, otherwise it will get an error. For example:

# Use only single node 
#SBATCH -N 1