Setting up VSCode for Rackham and Snowy¶
Info
- You can run VSCode on your local and still be able to work with modules loaded or environment created on Rackham.
- Similarly it is possible to take advantage of Snowy GPUs meanwhile developing on your local computer.
1. Connect your local VSCode to VSCode server running on Rackham¶
A step by step approach is available on UPPMAX VSCode documentation.
When you first establish the ssh connection to Rackham, your VSCode server directory .vscode-server will be created in your home folder /home/[username].
2. Install and manage Extensions on remote VSCode server¶
By default all the VSCode extensions will get installed on your home folder /home/[username]. Due to less storage quota on home folder 32 GB, 300k files, can quickly fill up with extensions and other file operations. The default installtion path for VSCode extensions can however be changed to your project folder which have way more storage space and file count capacity, 1TB, 1M files.
2.1. Manage Extensions¶
Go to Command Palette Ctrl+Shift+P or F1. Search for Remote-SSH: Settings and then go to Remote.SSH: Server Install Path. Add Item as remote host rackham.uppmax.uu.se and Value as folder in which you want to install all your data and extensions /proj/uppmax202x-x-xx/nobackup (without a trailing slash /).
If you already had your vscode-server running and storing extensions in home directory. Make sure to kill the server by selecting Remote-SSH: KIll VS Code Server on Host on Command Palette and deleting the .vscode-server directory in your home folder.
2.2. Install Extensions¶
You can sync all your local VSCode extensions to the remote server after you are connected with VSCode server on Rackham by searching for Remote: Install Local Extensions in 'SSH: rackham.uppmax.uu.se' in Command Palette. You can alternatively, go to Extensions tab and select each individually.
2.3. Selecting Kernels¶
Request allocation in either Rackham or Snowy compute node depending on your need, for that use interactive slurm command. Load the correct module on Rackham/Snowy that you contains the interpret you want on your VSCode. For example in case you need ML packages and python interpreter, do module load python_ML_packages. Check the file path for python interpreter by checking which python and copy this path. Go to Command Palette Ctrl+Shift+P or F1 on your local VSCode. Search for "interpreter" for python, then paste the path of your interpreter/kernel.
venv or conda environments are also visible on VSCode when you select interpreter/kernel for python or jupyter server. For jupyter, you need to start the server first, check Point 3.
3. Working with jupyter server on Rackham and snowy¶
Rackham:¶
Module load jupyter packages either from module load python or module load python_ML_packages as per your needs. For heavy compute and longer running jupyter server, allocate a Rackham compute node instead of using login node. Either request for rackham compute node by using, for example, interactive -A uppmax202x-x-xx -p node -N 1 -t 2:00:00 or move to the next step to run jupyter on login node itself. Start the jupyter server jupyter notebook --ip 0.0.0.0 --no-browser. Copy the jupyter server URL which goes somethig like http://r52.uppmax.uu.se:8888/tree?token=xxx, click on Select Kernel on VSCode and select Existing Jupyter Server. Past the URL here and confirm your chioce.
Snowy:¶
Start an interactive session with GPU allocation on Snowy interactive -A uppmax202x-x-xx -p node -N 1 -t 02:00:00 --gres=gpu:1 -M snowy. Module load the jupyter packages module load python_ML_packages and start the jupyter server jupyter notebook --ip 0.0.0.0 --no-browser. This should start a jupyter server on Snowy compute node with one T4 GPU. Copy the URL of the running jupyter server which goes something like http://s193.uppmax.uu.se:8888/tree?token=xxx and paste it in the jupyter kernel path on your local VSCode. The application will automatically perform port forwarding to Rackham, which already is listening to Snowy compute nodes over certain ports.