Kai Schlamp

Sorry, I am still a bit confused 😞. So with #1186 we will do it the other way around than > there would always be a path where you upgrade...

> It is confusing :( When we wrote procrastinate, we were in an environment where we couldn't stop the service while applying migrations, so we had to ensure we created...

I am a bit hesitant to add this, as it ultimately just translates to a string and only deludes that it does something special. @ewjoachim What do you think?

Some ideas @ewjoachim and I had after some discussion (so that we don't forget about it): - In general, when a custom encoder is set, encode the args with the...

> * A/ Do nothing, and let this length problem be > * B/ Find something smart to do about it, which we haven't found yet > * C/ Fix...

If I understand it correctly, the Procrastinate worker uses its own connection pool independent of the Django connection pool (it just fetches the connection parameters from Django). But I wonder...

> Very possibly! But now I’m confused. Are you saying the task itself is executed using a Django connection, rather than a procrastinate one? Yes, I think so. Procrastinate [extracts...

@ewjoachim When [Procrastinate fetches the connection parameters](https://github.com/procrastinate-org/procrastinate/blob/787aa3bcf965bdee092f8238f15916b538197baf/procrastinate/contrib/django/utils.py#L26) from Django it [creates a new (Django) connection](https://github.com/django/django/blob/aa5293068782dfa2d2173c75c8477f58a9989942/django/utils/connection.py#L62). I wonder if we should explicitly close it after we got the parameters. I don't...

I dug a bit deeper, but I was wrong. When the connection parameters are fetched from Django, no connection is opened. Contrary to the name `create_connection,` it just creates a...

> Huh! Ok. Weird. The pool in Django I thought would do that. It's a new feature though - maybe it is a bit buggy on there end. Would there...