BentoML

What's the `_result_store` ( a `Sqlite3Store`) on a `ServiceAppFactory`  ?
Or more generally, why is a BentoML storing data locally in a sqlite database ? Is this needed to run a service, can we disable it ?

This is only used when using async tasks <https://docs.bentoml.com/en/latest/get-started/async-task-queues.html|https://docs.bentoml.com/en/latest/get-started/async-task-queues.html>

Ok, we are getting sporadic `sqlite3.OperationalError: database is locked`  on the `CREATE TABLE IF NOT EXISTS result ...`  query.
I'm guessing this is a race condition when the service is created in multiple processes. As they all init the service, `Sqlite3Store` instances are created and entered (`__aenter__` is called) at _the same time_. So two processes try to create the DB file and/or the table, and one of them is kicked out by the lock.

Could you share a bit more about your setup? Do you run multiple BentoML serve instances on the same host? 

We built a container image with `bentoml containerize` and deploy the produced image on a kubernetes cluster.

We do run multiple containers n the same host, but they should each get their own filesystem.
We do run two workers in each container, so as I understand it we do have two processes creating their own service instance.

okay, I will move the init_db to `on_deployment` to avoid race condition

Hello <@U08EDB1DH5M>, thank you for helping us to spot this. We have released a PR to fix this, <https://github.com/bentoml/BentoML/pull/5369>

once it is merged, you can install from the `main` branch to test it out. We will issue a new release very promptly.
Meanwhile, unfortunately you will have to limit the workers to 1 for now.

FYI bentoml v1.4.15 is released with the fixes