Skip to content

Latest commit

 

History

History
165 lines (119 loc) · 6.21 KB

vscode.rst

File metadata and controls

165 lines (119 loc) · 6.21 KB

VSCode on Triton

VSCode is a text editor and IDE (Integrated Development Environment). It is very popular these days, partly due to it's good usability.

Installation

If you are using on Triton, it's available as a web app through :doc:`Open OnDemand <../usage/ood>`, see below.

It can also be installed on your own computer, which might be good to do anyway. If you do this, make sure you turn off telemetry if you don't want Microsoft to get reports of your activity. Search "telemetry" in settings to check and disable (note that this doesn't fully turn it off).

VSCodium is an open-source build of VScode (like "chromium" is an open-source build of Google Chrome) that disables all telemetry by default and removes non-open source bits. It is essentially the same thing, but due to Microsoft licenses it can't use the same extension registry as VSCode. It does have a stand-alone extension registry, though.

Security and extensions

As always when using user-contributed extensions, be cautious of what extensions you install. A malicious extension can access and/or delete all of the data available via your account.

VSCode through Open OnDemand

.. seealso:: :doc:`../usage/ood`

VSCode is available through :doc:`../usage/ood`, and with this you can select whatever resources you want (memory, CPU, etc) and run directly in the Slurm queue. This means you can directly perform calculations in that VSCode session and it runs properly (not on the login node).

This is useful for getting things done quickly or do simple debugging, but running in a web browser can be limited in some cases (interface, lifetime, etc.). Another limitation is OOD VSCode does not support GPU computing.

VSCode remote SSH

"Remote SSH" is a nice way to work on a remote computer and provides both editing and shell access, but everything will run directly on the login node on Triton. This is OK for wiring code and editing, but not for main computations (see the section above or below). To repeat: don't use this for running any big computations.

Screenshot saying "SSH: triton".

If you see this in the lower left corner (or whatever the name of your cluster SSH config is), you are connected to the login node (and again, should not do big calculations). It's possible the exact look may be different for others.

You can see connection instructions (including screenshots) at the Sigma2 instructions.

VSCode can use a regular OpenSSH configuration file, so you may as well set that up once and it can be used for everything - see :doc:`/scicomp/ssh` for the full story. The basics of SSH to Triton are in :ref:`triton-connecting-ssh`. A SSH key can allow you to connect without entering a password every time.

VSCode remote SSH host directly to interactive job

Sometimes you want more resources than the login node. This section presents a way to have VSCode directly connect to a job resource allocation on Triton - so you can do larger calculations / use more memory / etc. without interfering with others. Note that for real production calculations, you should use :doc:`../tut/serial`, and not run stuff through your editor, since everything gets lost when your connection dies.

This section contains original research and may not fully work, and may only work on Linux/Mac right now (but Windows might work too since it uses OpenSSH).

In your ~/.ssh/config, add this block to define a server triton-vscode. For more information .ssh/config, including what these mean and what else you might need in here, see :doc:`/scicomp/ssh`:

Host triton-vscode
    ProxyCommand ssh triton /share/apps/ssh-node-proxycommand --partition=interactive --time=1:00:00
    StrictHostKeyChecking no
    UserKnownHostsFile /dev/null
    User USERNAME

# You also need a triton alias here:

Host triton
    HostName triton.aalto.fi
    # ... any other thing you need for connecting to triton.
    User USERNAME

Now, with VSCode's Remote SSH, you can select the triton-vscode remote. It will ssh to Triton, request a job, and then directly connect to the job. Configure the job requirements in the ProxyCommand line (see :ref:`ref-job-submission` - you can have multiple Host sections for different types of requirements).

Possible issues which may affect usage:

  • If the ssh connection dies, the background job will be terminated. You will lose your state and not be able to save.
  • If the job dies due to time or memory exceeded, the same as above will happen: your job will die and there is no time to save.
  • If you srun from within the job, then it gets messed up because the environment variable SLURM_JOB_ID is set from the interactive job that got started. It's hard for us to unset this, so if you are using the terminal to srun or sbatch, you should unset SLURM_JOB_ID first. (Note there are many other variables set by Slurm. Make sure that they don't interfere with jobs you may run from this vscode session).

Sometimes you want to debug your code using GPU(s). To connect to a GPU node, you can follow the above guide, and request a gpudev partition. Note that gpudev partition has a strict time limitation: Every session is maximum 30 minutes long. For more information, see :doc:`partitions <../ref/partitions.rst>`.

Host triton-gpu-debug
ProxyCommand ssh triton /share/apps/ssh-node-proxycommand --partition=gpudev --time=0:30:00 StrictHostKeyChecking no UserKnownHostsFile /dev/null User USERNAME

# You also need a triton alias here:

Host triton
HostName triton.aalto.fi # ... any other thing you need for connecting to triton. User USERNAME

VSCode debugging

Following this guide, you can attach a debugger to a running job.