JANELIA_REMOTE_DEBUG.md

How to use remote debugging over SSH

Following instructions from: https://code.visualstudio.com/docs/python/debugging#_remote-debugging

Launch the remote debugging server

Reserve a slot and start up the remote debugger server, with the program arguments. (And make note of the cluster node you end up on.)

ssh login1

# From login1, Launch an interactive LSF job (in this example, I use 4 slots)
bsub -n 4 -q interactive -Is /bin/bash

# Activate your conda environment
conda activate my-env

# Install the Visual Studio remote debugger
conda install -c conda-forge debugpy

echo "Starting remote debugging server on $(uname -n)"
python3 -m debugpy --listen 0.0.0.0:3000 --wait-for-client $(which flyem-snapshot) -c my-config.yaml

VSCode debugging configuration

In VSCode, edit launch.json to add the following to the "configurations" list:

{
    "name": "Python: Remote Attach",
    "type": "python",
    "request": "attach",
    "connect": {
        "host": "MY_REMOTE_MACHINE.int.janelia.org",
        "port": 3000
    },
    "pathMappings": [
        {
            "localRoot": "/Users/bergs/workspace/flyem-snapshot",
            "remoteRoot": "/groups/flyem/proj/cluster/miniforge/workspace/flyem-snapshot"
        }
    ]
},

...but replace MY_REMOTE_MACHINE with the name of the remote node (e.g. something like h06u01), and replace the values for "localRoot" and "remoteRoot" to match the path of the git repo on the local and remote machines.

After the debugging server is launched, select your new configuration from the VSCode debugging panel and hit 'run'.

ssh tunnel

If the chosen port (3000) can't be directly accessed from your local laptop due to firewall rules, then start up an ssh tunnel by executing a command like this one on your local machine (your laptop):

$ ssh -L 3000:${REMOTE_MACHINE}:3000 bergs@login1

Then, replace MY_REMOTE_MACHINE with "localhost".