RSP was working great for me for a while. Recently (today?) I noticed that when I try to run a cell, it often shows the “in progress” mark [*] for a while – seconds to minutes – before executing the cell, even if the command is quick, say a print. Once the cell does eventually run, it lists a time that I’d expect for the cell’s runtime (1-10 ms) – much shorter than the wall time from running the cell to the finish. The problem persisted even after restarting my server and changing my server size from large to medium. Does anyone know what causes this, or have advice about how to avoid it?
Hmm. In your home space you will find a director called ‘system-test’. Inside is a notebook called token-info with one big cell. Run it once to pick up the imports, and then run it a second time and see if it has the delay you have seen?
(Nothing special about that notebook, it’s just that it is mostly prints and I can compare it with what I see)
I gave that notebook a bunch of runs, and it worked great/quickly. I’ve found that the hanging/delay is sporadic – I haven’t experienced it for a few days. I wonder if it’s a usage thing, or whether the notebook sets the “in progress” mark locally and this could be a connectivity issue?
Thursday evening is our node upgrade window for Google - we can’t time it exactly in the patch window we have to give then a 12-hour window - so things can get a bit unpredictable on the cluster. I note that’s when you filed this report so maybe that? Feel free to open an RSP support issue here noting any further occurences, the more data we have the better.
We are planning on adding timing data to our testing harness which will also detect issues like this.
This makes a lot of sense to me, especially with how it appeared to be a transient issue.