Interactive usage

When you login to CSC supercomputers, you end up in one of the login nodes of the computer. These login nodes are shared by all users and they are not intended for heavy computing. See our Usage policy for details. If you need to do heavy computing interactively, you should use interactive batch jobs.

In an interactive batch job, a user submits a batch job that starts an interactive shell session in a computing node. For heavy interactive tasks users can also request specific resources (time, memory, cores, disk). You can also use tools with graphical user interfaces in this interactive shell session, but in this case you need to have enabled X11 forwarding. An even better option for remote graphics is to use the Puhti web interface.

Please notice that the interactive batch jobs run in the computing nodes, where the environment differs slightly from the login nodes. For example, not all the same text editors are available. Furthermore, when you log out from an interactive batch job, the session with all the processes will be terminated, and data in the job specific $TMPDIR and $LOCAL_SCRATCH areas will be removed.

Easy interactive work: sinteractive command

Puhti and Mahti have an interactive partition which enables immediate access to an interactive batch job session. The easiest way to use this resource is to use the sinteractive command:

sinteractive -i

The command above asks what computing project will be used and how much resources the job will need. After that it opens a shell session that runs on a compute node. You can use this session as a normal bash shell without additional Slurm commands for starting jobs and applications.

You can define the resource requests in command line too if you don't want to specify them interactively. Note that the sinteractive commands in Puhti and Mahti are not identical. There is some differences in both command line options and in the way how the command works.

sinteractive in Puhti

In Puhti, each user can have two active sessions open in the interactive partition. In the interactive partition you can reserve in maximum 8 cores, with max 76 GB of memory, up to 7 days of time, and 720 GB of local scratch space. GPUs cannot be reserved.

If your requests exceed these limits or you already have two sessions in the interactive partition, sinteractive can submit the session request to small or gpu partitions instead. However, in these cases your session starts queuing just like normal batch job and you may need to wait some time before the requested resources become available and the interactive session starts.

All the sinterative sessions are executed in nodes that have NVMe fast local disk area available. The environment variable $TMPDIR points to the local disk area of the job. This local disk area has high I/O capacity and thus it is the ideal location for temporary files created by the application. Note however, that this disk area is erased when the interactive batch job session ends.

For example, an interactive session with 2 cores, 8 GiB of memory, 48 h running time and 100 GiB local scratch using project project_2001234 can be launched with command:

sinteractive --account project_2001234 --cores 2 --time 48:00:00 --mem 8000 --tmp 100

Available options for sinteractive in Puhti are:

Option	Function	Default	Max
-i, --interactive	Set resource requests for the job interactively.
-t, --time	Run time reservation in minutes or in format d-hh:mm:ss.	24:00:00	7-00:00:00
-m, --mem	Memory reservation in MB.	2000	76000
-j, --jobname	Job name.	interactive
-c, --cores	Number of cores.	1	8
-A, --account	Accounting project.
-d, --tmp	Size of job specific $TMPDIR disk (in GiB).	32	720
-g, --gpu	Number of GPU:s to reserve (max 4)	0	0

sinteractive in Mahti

In Mahti, users can have up to 8 interactive batch job sessions in the interactive partition. Other partitions don't support interactive batch jobs. Each interactive session can reserve 1-32 cores, but the total number of reserved cores shared with all user sessions cannot exceed 32. Thus a user can have for example 4 interactive sessions with 8 cores or one 32 core session. Each core reserved will provide 1.875 GB of memory and the only way to increase the memory reservation is to increase the number of cores reserved. The maximum memory, provided by 32 cores, is 60 GB.

For example, an interactive session with 6 cores, 11,25 GiB of memory and 48 h running time using project project_2001234 can be launched with command:

sinteractive --account project_2001234 --time 48:00:00 --cores 6

Available options for sinteractive in Mahti are:

Option	Function	Default
-i, --interactive	Set resource requests for the job interactively
-t, --time	Run time reservation in minutes or in format d-hh:mm:ss.	24:00:00
-j, --jobname	Job name.	interactive
-c, --cores	Number of cores ( + 1.875 GB of memory/core)	2
-A, --account	Accounting project.

Example: Running a Jupyter notebook or RStudio server via sinteractive

See the Using RStudio or Jupyter notebook tutorial.

Example: Running an MPI job in an interactive session

Since the shell started in the interactive session is already a job step in Slurm, more job steps can't be started. This will disable, e.g. running GROMACS tools, as gmx_mpi is a parallel program and normally needs srun. In this case, in the interactive shell, srun must be replaced with orterun -n 1. Orterun does not know of the Slurm flags, so it needs to be told how many tasks/threads to use. The following example will run a GROMACS mean square displacement analysis for an existing trajectory.

sinteractive --account <project>
module load gromacs-env
orterun -n 1 gmx_mpi msd -n index.ndx -f traj.xtc -s topol.tpr

To use all requested cores in parallel, you need to add --oversubscribe. E.g. for 4 cores, a parallel interactive job (launched from the interactive session) can be run with

sinteractive --account <project> -c 4
module load gromacs-env
orterun -n 4 --oversubscribe gmx_mpi mdrun -s topol.tpr

Explicit interactive shell without X11 graphics

If you don't want to use the sinteractive wrapper, it's possible to use Slurm commands explicitly. Note, as you may need to queue, it's convenient to ask for an email once the resources have been granted.

srun --ntasks=1 --time=00:10:00 --mem=1G --pty \
  --account=<project> --partition=small --mail-type=BEGIN \
   bash

Once the resource has been granted, you can work normally in the shell. The bash prompt will show the name of the compute node:

[csc-username@r07c02 ~]$

Once the requested time has passed, the shell exits automatically.

Starting an interactive application with X11 graphics

To enable X11 graphics, add --x11=first to the command. The following will start the application myprog:

srun --ntasks=1 --time=00:10:00 --mem=1G --x11=first --pty \
  --account=<project> --partition=small --mail-type=BEGIN \
   myprog

Note, that you can replace myprog with bash, which will launch a terminal on the compute node. From that terminal you can launch also graphical applications. Once the requested time has passed, the application will be automatically shutdown.

Do you find this page useful?

Yes

Thank you for your feedback!

No

Thank you! Your feedback is appreciated.

Please let us know how we could improve this page by contacting us or by creating an issue in GitHub. You can also edit this page yourself and contribute your improvements by creating a pull request.