aaw

aaw copied to clipboard

Describe the bug

Using VS Code in VNC notebook servers we found that it often crashes due to unknown reasons, not providing any error message. Crashes frequently, on multiple days and observed on two notebook servers.

Environment info

Namespace: eventdetec

Notebook/server: "mlflow"

Steps to reproduce

Start VS Code via terminal code . and the crash will appear sometime. We cannot provide more details as the cause triggering the crash is unknown.

Expected behaviour

No crash, VS Code works normal.

Screenshots

If applicable, add screenshots to help explain your problem.

Additional context

Add any other context about the problem here.

How often does this occur @chritter ? It will probably be difficult to reproduce this issue, however if it does occur reliably for you, perhaps you can create a crash dump for us to use to investigate?

Instructions: https://github.com/Microsoft/vscode/wiki/Native-Crash-Issues

Here is a screenshot just now. As you see VS Code is not rendering properly.

It happens frequently, also reported by my colleague Krishna.

So it continues to run but the functionality / interface is broken?

If it doesn't actually trigger a crash then the dump file probably doesn't get created... so maybe that won't be helpful.

I was able to produce some similar issues today (but not very frequently) and produced a dump file, however after processing it didn't indicate a specific culprit.

Some common suggestions are to try adding the following flags to your vscode shortcut: --disable-gpu --disable-extensions

Obviously disabling extensions makes vscode much less useful so I would only recommend testing with that one very briefly to see if it helps indicate an issue with an extension. The disable GPU, if not needed, could be a solution.

I would also suggest that you ensure your notebook has enough resources, to eliminate that as a concern.

Lastly I'll suggest updating the version of vs code that we package as a newer version could always help! :)

@vexingly We will test your suggestions of the flags. I am waiting for resource monitoring via Grafana to be enabled again to see if some other services overutilize cpu.