-
Notifications
You must be signed in to change notification settings - Fork 94
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
[Triage] WebSocket Errors on AML compute instance with Jupyter Notebooks #1386
Comments
I don't know of anything in this area on vscode's end that changed in this release, @SiddhanthUnnithan any idea? |
Thanks @roblourens. Perhaps nothing changed; however, the experience is less than stellar. When does @alexdima return frm vacation? |
Thanks for letting us know @YnnamTenob -- adding @tbombach to this thread to help triage. |
started again a few ahours ago. no more than one minute after connecting the WebSocket errors out erverStartTime==== |
I have been having the same issue. Please let us know what I can share to provide more details on this. |
@YnnamTenob Looking at logs shared in #1234 , it appears to me that the remote extension host becomes unresponsive. Could you perhaps try to use extension bisect to see if one of your extensions is causing very high CPU usage of the remote extension host?
|
@alexdima I'm not sure how this might work when bisect starts it disables everything I need to do work to detect the problem so I can't determine if the problem exists because I cannot run the notebook. There is no way for me to say if its good or bad. I said it was bad bisect said it was done and it was a problem in code. |
@YnnamTenob Could you perhaps try to reduce the extensions you are using to the bare minimum manually using Disable Extension? I understand that you cannot do what you intended to, but by disabling the extensions and then doing something simple, like editing a plain text file, it would help troubleshoot where a problem is occurring. When I have seen this in the past, it was often the case that one of the extensions would run up the extension host process to 100% CPU, which would trigger our UI to believe connection was lost to the remote extension host. (could you perhaps take a look at |
@alexdima when I am connected to the Azure ML Remote all extensionsions are disabled with the exception of: I think I can only turn off Sandance and Azure Cognitive Search I need all the rest to do my work. I have done that. |
Yesterday was a good day VSCode behaved well did not give me any socket issues and did not hog the memory - cpu. Today VSCode is once again up to no good. I have determined that the problem is in how VSCode reacts when it's websockets connections are failing in this remote mode. VSCode is not able to gracefully cope with this and in turn makes development difficult if not impossible if the connection to Azure is weak or intermittent. I don't know if the intermittent behavior is caused by Azure itself or if it is a function of my network connectivity. I am always connected via gigabit fibre but that is not sufficient evidence that my network is not the issue. In a nutshell I think you guys need to rethink how VSCode behaves under these conditions. I think I have a solid connection but in any case VSCode should not go into fits if the connection fails on either end. |
I had the same experience so far. The problem occurs due to the failing websockets in the remote mode. Though I got to mention that my network connection was not stable, though development in Jupyter notebook in the meantime was smooth and developing in VSCode was not possible for me. |
Thank you for following up. @sevillal Given the information we have at this time, I don't believe this to be an issue with VS Code core, but with the the connection created by the |
Hi, Same problem here. I have also the same situation than @eduedix mentions. Connection is amazing and jupyterlab works fine but VScode does not. BR |
I started to get websocket errors again with stable internet connection.
I guess this has to do with the amount of data being transferred via the socket. The failures occur when I create plots with many data points in vscode notebook. You can try it out by line charting 500k datapoints. |
@alexdima transfered to vscode-tools-for-ai |
@eduedix do you have a sample notebook to repro the issue? Thanks |
Closing this as there has been no response for over 7 days. Please reopen if need be. |
Issue Type: Bug
After the July update 4 days ago connections to compute instances become unstable very quickly. VSCode reports socket Errors.
[Error - 2021-08-08 22:49:35.458] Websocket 6 error sending data: Error: WebSocket is not open: readyState 3 (CLOSED)
[Error - 2021-08-08 22:49:41.032] Websocket 7 error sending data: Error: WebSocket is not open: readyState 3 (CLOSED)
These errors appear to render VSCode unusable du to slowness. during this episode VSCode exhibits High Memory and CPU usage. The only way to get VSCode working again is to esstablish a new connection with the remote compute resource; however after only a few minutes that connection experiences the same error.
VS Code version: Code 1.59.0 (379476f0e13988d90fab105c5c19e7abc8b1dea8, 2021-08-04T23:13:12.822Z)
OS version: Windows_NT x64 10.0.17134
Restricted Mode: No
Remote OS version: Linux x64 5.4.0-1043-azure
System Info
gpu_compositing: enabled
multiple_raster_threads: enabled_on
oop_rasterization: enabled
opengl: enabled_on
rasterization: enabled
skia_renderer: enabled_on
video_decode: enabled
vulkan: disabled_off
webgl: enabled
webgl2: enabled
Extensions (9)
A/B Experiments
The text was updated successfully, but these errors were encountered: