Prerequisities

  • Science (CNCZ) account.
  • Being member of the qm, qmorgchem or qmteaching and gaussian groups (ask Jan for access).
  • Vpn (when working from home).
  • SSH client (linux usually has OpenSSH, windows - for instance MobaXterm or Putty).
  • X11 client (linux usually has OpenSSH, windows - for instance MobaXterm
  • (optional) WinSCP

Video guides

GUIDE how to setup environment for Gaussian calculations in windows.

Setting-up the environment (to be done once)

To submit calculations easily you need to set-up the environment first.

Automated script

You must be within university network or connected through VPN. connect through SSH to lilo.science.ru.nl (in putty that is the hostname). Your username is your science login. After logging-in, execute /vol/qmorgchem/scripts/gaussify script (cnorris is the username in this example):

cnorris@lilo7:~$ /vol/qmorgchem/scripts/gaussify

If you’re more experienced user, you can setup the things on your own.

Explanation of the automated script

The bash script sets-up few things on your behalf. What it does is:

~/.profile file editation

to see your ~/.profile content, execute:

cnorris@lilo7:~$ cat ~/.profile

you should see something like this:

export PATH=$PATH:/vol/qmorgchem/scripts
module load Gaussian-16
umask 027

This file stores some of your server settings, the script:

  • Adds (export PATH=… line) /vol/qmorgchem/scripts in your $PATH variable - by doing this you can easily execute scripts located in /vol/qmorgchem/scripts folder such as the important G16 script.
  • Automatically loads Gaussian-16 module (module load… line) which enables you to use commands like cubegen and similar.
  • Changes the umask to 027 - that means that files newly created from the console can be readed by people from qmorgchem, but cannot be changed by them. If you want to keep the files changeable by other people from qmorgchem, then remove this line from your ~/.profile file.

Folders and shortcuts creation

Script creates your folder in qmorgchem disk. If you are member of the qmorgchem group it creates your folder also on the qmorgchem-old disk. It creates shortcuts (named symlinks in linux) to those folders within your $HOME directory by ln -s commands.

To see those folder changes, execute:

cnorris@lilo7:/vol$ ls -lh ~

You should then see something like this:

total 24K
drwxr-s--- 2 cnorris cnorris 4.0K Apr 1 13:37 '$RECYCLE.BIN'
-rw-r----- 1 cnorris cnorris 402 Apr 1 13:37 desktop.ini
lrwxrwxrwx 1 cnorris cnorris 28 Apr 1 13:37 old -> /vol/qmorgchem-old/cnorris/
lrwxrwxrwx 1 cnorris cnorris 15 Apr 1 13:37 others -> /vol/qmorgchem/
lrwxrwxrwx 1 cnorris cnorris 24 Apr 1 13:37 running -> /vol/qmorgchem/cnorris/

ls is a command which lists directory content, ~ is a shorthand for your $HOME directory (/home/cnorris in case of cnorris user), -lh are arguments: l stands for “long” and h stands for “human-readable.” You clearly see that shortcut named “old” points to your folder at qmorgchem-old disk and “running” points to your folder at qmorgchem disk. Please use qmorgchem-old for inactive projects and do not store .chk files there.

Checking calculations, submitting jobs, checking queues

Queues

Each computational job must be submitted to a queue. Currently we have two queues - orgfast and orglong.

  • orgfast - Queue should be used for standard jobs (e.g., coptimization of geometries, harmonic frequency calculations, excited state spectra calculations). Please use max 4 cpus and maxi 12 GB of ram in this queue.
  • orglong - Queue for more demanding jobs (typically days or weeks of computational time). This queue has a lower priority; jobs will be running only if the given number of processors is free and not requested in the orgfast queue. Please use max 12 cpus and max 32 GB of ram in this queue.

Submitting jobs, checking queues

There are currently two queues available on the cluster: orgfast for small and quick jobs and orglong for bigger and more-demanding jobs. We are currently using Slurm as our job queuing manager.

To submit a job and doing anything with the running jobs you MUST be on cnlogin22

Basic Commands:

  • G16 $job_filename $queue - for instance G16 co2capcture.gjf orgfast - submits a calculation
  • squeue - checks queue
  • scancel $jobid - cancels a job
  • scontrol show jobid $jobid - show details of a job with job number $jobid (good for getting a full name)

Once calculation is running a .log file will appear in the folder where is the input file.

Putty

Connect to cnlogin22.science.ru.nl - details are the same as for the abovementioned for lilo.science.ru.nl. Change directory to your active directory where computations are being made (cd running - this brings you through shortcut to your folder on the qmorgchem disk).

To submit a job, type:

cnorris@cnlogin22:~$ G16 $job_filename $queue

command where $job_filename is filename of your Gaussian input and $queue is either orgfast or orglong.

WinSCP

connect through SFTP to cnlogin22.science.ru.nl and then open folder running and then type:

cnorris@cnlogin22:~$ G16 $job_filename $queue

command where $job_filename is filename of your Gaussian input and $queue is either orgfast or orglong.

For clarity, see the pictures below:

Winscp1.png|opening console Winscp2.png|logging in Winscp3.png|submitting

Checking calculations

once calculations are finished or you want to take a look, you can either use molden (its installed on the cluster - you can run it when you have mobaxterm on and you are connected on lilo) or you can connect to the disks and check the files from your windows molecule viewer.

Connecting to disks - windows

Open file explorer and into navigation bar type:

\\qmorgchem-srv.science.ru.nl\qmorgchem\

for qmorgchem-old, the link is:

\\qmorgchem-old-srv.science.ru.nl\qmorgchem-old\

Connecting to disks - Linux

create some file in your homedir - for example lilofs:

mkdir ~/lilofs

install sshfs if you do not have it yet and then:

sshfs lilo.science.ru.nl:/ ~/lilofs

the disks are then located in ~/lilofs/vol , I will leave creating appropriate symlinks as a homework excercise - its analogous to commands mentioned above.

Standard operation procedure, tips

  • If the orglong queue is full and you have quick jobs, run them at orgfast
  • Pay attention when setting-up memory allocation, use of excessive memory prevents other jobs from running. Memory is shared. Do not use memory that you do not need.
  • Do calculations in optimization steps
  • reading force constants (RCFC) speeds-up the initial optimization step.

optimization steps

This section is an example of the optimization steps. Basis sets and levels of theory are case-specific and thus may vary.

  • semiempirical (if applicable, AM1, PM6)
  • cheap DFT (pure DFT, fe: B97D/sto-3g, M06L/midix)
  • expensive DFT (hybrid, fe: B3Lyp)

Issues

Open

Failing calculations after submission

error note: galloc: could not allocate memory. For now there is an issue that sometimes calculation fail while submitting a job. If this happens to you, please record a job number and send it to Jan. Do not delete .error, .out and .log files of the failed job, they are important for debugging this issue.

Disclaimer

If you have any issue to report please write Jan.

This guide is a draft, if you have any suggestions how to improve the readability, understandability or any idea how to make it better, please do not be shy and write Jan.