Using a server

Unix basics
Commands
Servers
How to use the Digital Research Alliance of Canada servers
Author

Karen C. Goncalves dos Santos

Published

February 28, 2023

Connect to server : ssh

Secure shell protocol: will allow you to connect to a server and write commands into it.

  • ssh user@server.ca
    • After asking for the password, will log user to server.ca.
  • ssh -i myKey.ppk user@server.ca
    • The option -i means “identity file”, ssh will use this file as the password (you need to have it in your local computer and copy it to the server.
    • To create an identity file (also known as ssh-key) follow the instructions here

You can speedup the process of connecting by creating a config.ini file in your .ssh folder.

Let’s say your username is USER1 and your private key is stored in the .ssh folder as “myKey.ppk”. Copy the script below in your terminal, replacing USER1 with your username.

echo "Host narval rorqual fir nibi
    HostName %h.alliancecan.ca
    User USER1
    IdentityFile ~/.ssh/myKey.ppk" > ~/.ssh/config

echo does what it sounds like it does. It repeats what you just typed (inside the quotes, but they are not required for echo). Instead of printing the same text in the command line, it saves to the file ~/.ssh/config.

Now, you can login any of the 4 hosts by just typing the name of the host server: ssh narval or ssh rorqual, etc.

Note: Rorqual is a bit weird with the ssh-key, accepting it when you use it in a ssh session in MobaXterm, but not when you type it in the ssh command. You may still need to use the password if using that server.

Commands specific to the server

Softwares

In the server, you do not need to install most softwares, as they are already installed and are called modules. These modules are what you need to load to your session to be able to use the software.

Available softwares

  • module avail
    • Shows the softwares that are installed (available) in the server.
  • module spider
    • module spider softw shows at all the modules that are installed which contain the words ‘softw’. Will show a list of all the versions of those softwares.
    • module spider software1/1.0 will show a description for software1/1.0, if it is installed, and how to load it to your session in the server.
  • module keyword softw
    • Similar to module spider, it will search for the word softw among the installed modules, but the search extends to the description of the software. This is useful when you are searching for a software that does something specific but you don’t need one specific.
  • module load software1
    • If software1 is installed in the server AND you have loaded all the softwares you need for software1 to work, will load the default version of software1 to your session.
    • If you want an specific version of a software, you need to put it in the command (eg. module load software1/1.0)
    • If you have a version of a software loaded and you try to load another, the previous one will be unloaded, meaning you cannot load two versions of a software at the same time.
  • module save my_collection and module restore my_collection
    • If you use many softwares frequently, you can create collections that will allow you to load them all together. For example, the code below creates a collection with the modules you need to run blast, then saves it to the collection named “blast”:
    module load StdEnv/2020  gcc/9.3.0  blast+/2.12.0
module save blast
    • Now, every time you want to run blast, you just need to run the code below and then add your script for blast!
    module restore blast

sbatch

This command will add your script to a queue in the server to run when the required resourcers (time, memory, number of central processing units – CPUs) are available. You cannot be in your ${HOME} directory to run it or in folders inside it. You need to move to ${SCRATCH} or ${PROJECT} to be able to run.

  • sbatch myScript.sh
    • Will add myScript.sh to the queue
    • The script must start with the following lines:
      • #!/bin/bash
        • Specifies the software (bash) that will be able to read this script (we give it’s absolute path, if you type ls /bin in the server, you will find the file ‘bash’)
      • #SBATCH --time=00-00:00:00
        • Normally anything that comes after # is not read by the computer, but since it is followed by ‘SBATCH’, the software sbatch will read it and interpret that you are asking for this amount of time in the queue.
    • If nothing else is given, the software sbatch will allocated the minimum amount of memory and CPUs and will find the sponsor of your account to create the bill the use of the server in the name of the sponsor
      • If you have more than one sponsor (eg. laboidp and desgagne) there will be an error, because the server will not know who they should bill.
      • You can specify the sponsor in two ways:
        • Add the following to a new line in the beginning of your script: #SBATCH --acount=def-sponsor (replace “sponsor” by the sponsor of your account)
        • Add the name of the account to the sbatch command: sbatch --acount=def-sponsor myScript.sh

srun

Will add your script to the queue, it is as if you were running the script in the session, so until it has finished running you won’t be able to do anything else.

  • srun --account=def-sponsor --mem-per-cpu=16G --time=01:00:00 myScript.sh
    • Will run myScript.sh, billing the account def-sponsor, using a maximum of 16Gb of memory and for maximum 1h.
  • srun --account=def-sponsor --mem-per-cpu=16G --time=01:00:00 -i
    • Same as above, however there is no script to run because we want to use the resources interactively (-i). This is useful if you want to test a script that you know may use a lot of memory. When you are finished, if the time has not run out, you can use Ctrl + D or exit to go back to your session.

salloc

Similar to using srun -i, will allow the interactive use of the resources allocated. When you are finished, if the time has not run out, you can use Ctrl + D or exit to go back to your session.

squeue

Will show the queue of jobs waiting to run in the server.

  • squeue
    • Will print ALL the jobs in the queue, submitted by every user of the server, which are running or waiting to run.
  • squeue -u user or sq
    • If your username is user, the two commands are the same. Will print all your jobs that are running or waiting to run
    • Below is an example of what the result of squeue -u smithj would look like
JOBID USER ACCOUNT NAME ST TIME_LEFT NODES CPUS GRES MIN_MEM NODELIST (REASON)
123456 smithj def-smithj simple_j R 0:03 1 1 (null) 4G cdr234 (None)
123457 smithj def-smithj bigger_j PD 2-00:00:00 1 16 (null) 16G (Priority)
  • JOBID: each time you using sbatch, srun or salloc, you create a job and it is assigned an ID, you can use this ID to get information about the job.
  • Name: Unless you specify a different name (by using --jobname=someName or -j someName) it will be the name of the script submitted
  • ST: status of the job.
    • R = running
    • PD = pending, meaning the resources needed are not available so you need to wait
    • CG = completing
    • CA = cancelled
    • See other status codes here
  • Nodes: basically a groupping of CPUs
  • MIN_MEM: memory requested to run the job
  • Nodelist (reason): When your job is running, you will get a list of the nodes it is using. If it is pending, you will get a reason as to why; if it is already running, the reason will be “none”

scancel

Cancel jobs that are pending or running. Use it with the id of one or more jobs. You are not allowed to cancel a job submitted by someone else.

diskusage_report

Prints a report of the use of the server by you in your home and scratch directories, as well as your directories in the project folders (each sponsor has a different one, so you may have files in different project folders). The report includes the space used and the number of files in each directory, as well as the limits of these directories.

Resources

AWK

Back to top