You signed in with another tab or window. Reload to refresh your session.You signed out in another tab or window. Reload to refresh your session.You switched accounts on another tab or window. Reload to refresh your session.Dismiss alert
I just tried to get somewhere around 40 students to generate an Xtext grammar at the same time and it crashed the server. As far as I can tell, the Xtext container and possibly even the Docker daemon got killed. The Xtext container would attempt to restart automatically, which would create extra load on the server, together with all clients continuing to poll for console updates.
It would be worth doing some performance testing to work out what we can do to improve this. It may be an issue to do with the polling for successful editor generation, which we should probably change to server-side events / websockets to improve efficiency.
The text was updated successfully, but these errors were encountered:
The number of polling requests can be the culprit here. It's helpful if we can look into the monitoring software to see what resource was used the most. I believe the current polling method is server-side events, or at least it looks like SSE from the browser side. Switching to web sockets can reduce some overheads.
Thanks, @aryaei2000 for looking at this. The current polling method is, I think, even worse: it's plain polling :-) The culprit is this recursive call here:
Instead, checkEditorReady should be replaced by a server-side events or websocket implementation where it's the server's responsibility rather than the client's to send updates whenever they're ready.
I just tried to get somewhere around 40 students to generate an Xtext grammar at the same time and it crashed the server. As far as I can tell, the Xtext container and possibly even the Docker daemon got killed. The Xtext container would attempt to restart automatically, which would create extra load on the server, together with all clients continuing to poll for console updates.
It would be worth doing some performance testing to work out what we can do to improve this. It may be an issue to do with the polling for successful editor generation, which we should probably change to server-side events / websockets to improve efficiency.
The text was updated successfully, but these errors were encountered: