search

rebus-org / Rebus.RabbitMq

:bus: RabbitMQ transport for Rebus
https://mookid.dk/category/rebus
Other
62 stars 44 forks source link

Work on removing polling from RabbitMQ transport #72

Closed zlepper closed 3 years ago

zlepper commented 3 years ago

Hopefully implementation for rebus-org/Rebus#917

I hope by starting this we can start some discussion about how to implement it. Quite a few of the tests are failing now due to timeouts, which I assume is because the Receive method never returns.

Rebus is MIT-licensed. The code submitted in this pull request needs to carry the MIT license too. By leaving this text in, I hereby acknowledge that the code submitted in the pull request has the MIT license and can be merged with the Rebus codebase.

zlepper commented 3 years ago

I just cannot make this work as I want, so i'm closing this.

mookid8000 commented 3 years ago

Sorry to hear about that 😐

zlepper commented 3 years ago

No worries :)

Rebus expects to be able to poll the transport in a lot of places, and suddenly having the transport just hanging until Rebus gets a response doesn't really work out that well. There also seems to be some pushing already, so it's not all bad :) I would just have loved to get rid of the polling entirely, we didn't have to waste cpu cycles waking to check all the time.

mookid8000 commented 3 years ago

(..) waste cpu cycles waking to check all the time

Yeah well, while it does of course consume a little bit of CPU, I don't think it's that bad.... the driver has some kind of blocking + timeout on its polling method, meaning that Rebus does not call it that often when running idle. If would of course have been ideal if messages could be pushed

zlepper commented 3 years ago

The problem is we have to decrease the polling interval to make it not take a while the first time an event comes in, in a while, at least in my experimentation.