[POC] Requeue tasks after ETA is met #7305
Draft
Add this suggestion to a batch that can be applied as a single commit.
This suggestion is invalid because no changes were made to the code.
Suggestions cannot be applied while the pull request is closed.
Suggestions cannot be applied while viewing a subset of changes.
Only one suggestion per line can be applied in a batch.
Add this suggestion to a batch that can be applied as a single commit.
Applying suggestions on deleted lines is not supported.
You must change the existing code in this line in order to create a valid suggestion.
Outdated suggestions cannot be applied.
This suggestion has been applied or marked resolved.
Suggestions cannot be applied from pending reviews.
Suggestions cannot be applied on multi-line comments.
Suggestions cannot be applied while the pull request is queued to merge.
Suggestion cannot be applied right now. Please check back later.
Fixes #6500
This PR is just a proof of concept as requested here. It is far away to be something complete but it illustrates the idea and addresses the problem.
Description
Workers using
--prefetch-multiplier 1
andacks_late
should never reserve more tasks than their concurrency level. In other words, they should reserve 1 task for every child worker.However, when a worker receives an ETA task, it holds the task in memory (unacknowledged) and increases the QoS to allow RabbitMQ to send a new task. So, the QoS of workers is:
When the ETA is met, the worker decreases the QoS but still holds the task until one child worker gets idle. This PR attempts to solve the problem by just queueing the task back to RabbitMQ at the same position it was so that other idle consumers can process it immediately.
POC
The
eta_poc
example has the instructions to test the behaviour and to compare it with the current implementation.