@kolaente Any update on the issue of tasks disappearing from Project 1 (local instance) when updated from version v0.24.2 stable to v0.24.1-185-89f78cd369 with Typesense enabled and after Reindexing?
If I disable Typesense, the tasks do come back and all works as normal.
Can I be testing something else?
No updates really, I was busy fixing other issues. Will take a look soon, but didn’t get any luck reproducing it last time.
It still does not work after a reindex? The Could not find a filter field named buckets in the schema might be unrelated to Typesense.
No worries, thank you for checking it.
My prod is on 0.24.2 and so is the stable. Workaround of disabling Typesense still works even in Unstanble. I will wait for the next major stable release.
In the mean time, I will run more detailed test of pulling down older version docker into a fresh staging setup and then upgrade from there. Maybe that way the bug will be reproducible.
I’ve pushed 9e39ef9397 and edf206aba6, please check with the next unstable build if that fixes the problem (should be ready in ~45min). Please also do a full reindex.
@kolaente You are amazing! thank you! I will check unstable in 30min.
v0.24.1-213-edf206aba6docker exec vikunja /app/vikunja/vikunja indexProject 1 (that I have since day 1 of my instance) does not show all the tasks, only a handful of the tasks are displayed.
In the docker stdout, there are no errors, apart from this warning when I open Inbox project with no tasks in it:
2024-09-19T19:42:58.94851645Z: WARNING ▶ models/getRawTasksForProjects 1534 Unable to fetch tasks from Typesense, error was 'status: 404 response: {"message": "Could not find a field named `positions.view_52` in the schema for sorting."}'. Falling back to db.
And also, why would Stable have higher version number then Unstable? Am I doing something wrong?
Should I worry about this: 2024-09-19 19:55:49 1018 [Warning] Aborted connection 1018 to db: 'vikunja' user: 'ben' host: '172.18.0.5' (Got an error reading communication packets)
Lastly, Typesense’s logs file was 580MBs, so I flushed that, but it did not help. 
Turns out this was caused by the task position in Typesense not being saved as the correct field type the first time. This happened because the field is created per view on demand and its type is automatically inferred from the data saved. Now, when the first value for a particular position field is a float which could as well be an int (for example, 42.0), that field gets created as an int64 instead of float. Subsequent tries to save a float into that field will then fail. That would mean some tasks could still be indexed, those with a .0 position value, while others couldn’t.
Additionally, these errors were not logged in Vikunja and I was unable to reproduce this locally. I was able to reproduce this on my own instance though which helped fixing the problem.
I’ve pushed a fix in 50a0674835 and could not reproduce the problem anymore after that. Please check again with the current unstable build.
That’s caused by the way the version number is built. Bugs are fixed on the main branch first and then backported to a release branch (because features should not be included in a patch release) which then contains the tag. The main branch does not contain the tag.
The git hash (the text after the version number) is the important part.
@kolaente YES! It works! Thank you very much for this!
I am on Version: v0.24.1-221-207b88a286
Typesense is enabled in docker-compose.yml
I have reindexed my instance
And all projects and all tasks are showing up without workaround!
Whooo 
that was quite a ride
@kolaente I have just upgraded to Version: v0.24.3 in my production (stable) environment.
After upgrade, I ran docker exec <name of the Vikunja container> /app/vikunja/vikunja index and all the tasks are showing up!
Thank you for your help once again!