Closed tkrabel-db closed 8 months ago
We start an IO server, so we have a thread pool to consume messages from the FE
Perhaps this could be avoided if the server is fully async. Since now it works serially, I think it's not hard to encounter this kind of race conditions when the server is called from different threads.
In Databricks notebooks, we log LSP errors. One common class of errors we get is this.
This means that the cell document doesn't seem to exist (or is not retrievable for whatever reason) at the time the number of lines are computed.
We start an IO server, so we have a thread pool to consume messages from the FE. My guess is that we have race conditions, where one thread is deleting a cell while the other still tries to access it.