Locking needed?

In a previous version of my web app (using bokeh standalone server version <=3.3), I have seen issues (file corruption) when my users are simultaneously executing the pandas.DataFrame.to_csv call, which is used to save a small amount of state for an interaction with a specific subset of our large set of data.
And sorry, hard to design a simple example–the only test I have found is to have a number of colleagues hit our rather large app hard at the same time.
Unfortunately the nature of usage seems to mean that this is not a one-off.
Is this a know issue? Is there a sensible way to lock the pandas action?
Niels

Bokeh is asynchronous, but not multi-threaded, so I am not sure where the kind of locking you suggest would come into play? Are all user sessions trying to write to the same file?

No. Same file name, different working directory–so for some reason things get jumbled up.
Like you say–single thread, async: this shouldn’t happen. So I am barking up the wrong tree.
Saying this out loud, I probably should check that I specify the full path, rather than rely on a working directory, which could explain why it fails for a single thread.
Thanks for making sure I am looking in a more sensible place.

Just spitballing more: you only have one instance of the server running? If you were running multiple server instances behind a load balancer then that could potentially lead to issues (but, AFAIK, I only if the same path was used in multiple places).

No, I have only one instance running.
Thanks.

This topic was automatically closed 90 days after the last reply. New replies are no longer allowed.