Error at importing CSV with users

radinamatic commented 1 month ago

Observed behavior

I eventually managed to import the user's list from CSV, but in one session it was consistently giving me grief with this test case on Windows 10, even though I used the same CSV as in the last round of integration testing for 0.16 back in February. Steps I performed:

try to import a CSV file and got the error below
export the CSV in case something was different in the document structure (not that I could notice)
open the exported CSV in Excel, paste the data conserving the header structure, save the new CSV
try to import this new CSV, got the same error as with the original (previous) CSV
try to replicate in Ubuntu 24.05 with the 0.17b3 DEB asset installed, and Windows 11 with the 0.17b3 EXE: zilch, all imports well, no errors
restart the Windows 10 VM, and lo and behold, this time even the original CSV time imports correctly 🤷🏽‍♀️

This might be a fluke, so I'm not labeling it as a blocker, but might be worth investigating at some point.

Errors and logs

db-logs.zip

Expected behavior

Import users via CSV without issues.

Context

Kolibri version: 0.17b3
Operating system: Windows 10
Browser: Firefox

jredrejo commented 1 month ago

@radinamatic according to the traceback it looks like a locking problem with the database: i.e. it was open by another project and kolibri could not open it. Could it be a zombie process from a previous kolibri restart?

That explains the fix after rebooting the system: the process that was locking the db was not there anymore.

radinamatic commented 1 month ago

That makes sense @jredrejo, thank you! A zombie process could have been present from the previous test case I was doing on the same device.

I was wondering if the potential path forward would be to more gracefully react to this type of errors and direct the user to restart the device, which would be more useful than a 'Ooops!' message. 🙂

rtibbles commented 1 month ago

Possibly, although that would require strings to do so.

I think the better option here is to make how we run Kolibri on windows better to avoid zombie processes in the first place.

Either way, I don't think this should be in the 0.17 milestone.

learningequality / kolibri