Troubleshooting
This page lists common CS Bridge problems and their fixes. To look up an exact message, see Error messages.
Logs
Two logs show the cause of most problems; read them first.
| Log | Shows | Open it with |
|---|---|---|
| CS Bridge output | Each step the extension takes on your laptop, and each error | View → Output, then CS Bridge in the drop-down |
| Linkspan job log | What happened inside the job on the compute node | On the cluster: ~/.cybershuttle/logs/linkspan-session-<jobid>.out and .err |
To read the newest job log, log in to the cluster and run:
cd ~/.cybershuttle/logs && tail -n 50 "$(ls -t *.err | head -1)"
A missing job log means the job never started running. Check it with squeue -u $USER; a session's job is named
linkspan-session.
Symptoms and fixes
| Symptom | Cause | Fix |
|---|---|---|
| "Session validation failed" on Add | Slurm or the site's submit filter refused the job | Read Slurm's message; choose another account, partition or fewer resources |
| Stuck on Submitting… | The login node cannot download Linkspan from github.com | Check the output log; ask the site whether login nodes have outbound HTTPS |
| Stuck on Starting Linkspan… | The compute node cannot reach the Dev Tunnels relay | Read the job's .err log; compute nodes need outbound HTTPS to *.devtunnels.ms |
| Microsoft sign-in fails | The local network blocks login.microsoftonline.com or *.devtunnels.ms | Try another network, or ask your IT group |
No ptyHost heartbeat, or the remote window crashes | The job hit its memory limit, which killed the VS Code server | Start a session with more memory |
| The remote window disconnects | The Dev Tunnel connection dropped | Wait about a minute; if the card returns to Connect, click it |
| The card shows Unreachable | A status check or the connection to the node failed | Wait for the next check, or click Reconnect. For SSH host unreachable, see below |
| MFA is requested again | The shared SSH connection to the login node closed | Expected; answer in the SSH Authentication panel |
| Extensions missing in each new job | The VS Code server is installed afresh for each job | List them in the VS Code setting remote.SSH.defaultExtensions |
| Dev Containers errors in the remote window | The extension needs a container runtime the cluster lacks | Disable Dev Containers in that window |
| Slow file transfer | Traffic goes through Microsoft's relay | Use scp, rsync or Globus through the login node or a data transfer node |
SSH host unreachable means a background status check could not log in, because it never shows a login prompt. To log
in again, click Open Menu →
Create New SSH Session → the cluster. If the form appears without a prompt, click the draft card's refresh icon
(Refresh Slurm details). Answer the prompt and close the form; the card recovers at the next check.
Questions
How is this different from Remote-SSH alone? Remote-SSH connects to a host you can already reach. CS Bridge first gets you a compute node through Slurm and a path to it.
Why at least 2 CPUs and 4 GB? Less starves the VS Code server, and Slurm kills it.
Can someone else connect to my session? Not through SSH: the node's SSH server accepts only the session's key. On a shared node, see Credentials and data.
Unlisted problems
For a problem not listed here, search the issue tracker and open
an issue if none matches. Attach the CS Bridge output and the job's .out and .err files, and name the cluster.