Open greazer opened 2 years ago
Regarding the reconnection issue here is further explanation on the cause, the timestamps shown are just to depict an approximation of how events would happen:
To solve VS Code Server not running, we could install and start the server on each reconnection, but it won't resolve the reconnection because of the following:
If we start VS Code Server without --auto-shutdown, would that help in this case? What would be the cost/performance implications of having the server running indefinitely?
Once this condition is reached, will customer lose it's dirty files, editor state, etc?
Expected Behavior
Connection to running notebook on an Azure ML compute remains even after many hours
Actual Behavior
Disconnect occurs at some point with a modal dialog stating "Cannot reconnect. Please reload the window". This is coming from VS Code core (or remote services).
Steps to Reproduce the Problem
Per Rob on the VS Code team, this disconnection is expected. There may be ways of ensuring it doesn't happen via the Azure ML extension. This could be problematic for cost (though I'm not sure how it's different than running a kernel all night).
Barring the prevention of disconnect, VS Code core and the Jupyter extension team may need to develop a way to explicitly disconnect and reconnect to running kernels.