Hacker News Reader

their description of the problem seemed kind of obtuse, in practice, these connection-pool related issues have to do with 1. request is interrupted 2. exception is thrown 3. catch exception, return connection to pool, move on. The thing that has to be implemented is 2a. clean up the state of the connection when the interrupted exception is caught, then return to the pool.

that is, this seems like a very basic programming mistake and not some deep issue in Redis. the strange way it was described makes it seem like they're trying to conceal that a bit.

It's an open source library, I assume that logic is abstracted within it and that the "basic mistake" was one of the maintainer's.

I think most app using redis-py rarely cancel async redis command. Python async web frameworks is gaining popularity, but the majority of people using python for their web application is not using an async framework. And of those people that do use them, not many of them canceling async redis requests often enough to trigger the bug.

There is a 1 year old autoclosed issue which is very similar to OpenAI's issue: https://github.com/redis/redis-py/issues/2028