Skip to content

Remove the backoff delay from Geo repository sync

What does this MR do?

Remove the backoff delay from Geo::RepositorySyncWorker

Are there points in the code the reviewer needs to double check?

Why was this MR needed?

With the backoff delay, the worker - iterating every second - enqueues 25 jobs in a burst, which sit and do nothing for 5 minutes. This blocks the worker from doing anything else for that time. The jobs themselves are processed almost immediately once the timeout expires, leading to long periods of complete inactivity.

Keeping a low max_capacity while removing the backoff delay allows us to process jobs at a much higher rate whilst never going above the defined concurrency limit, so limiting load on the primary.

Screenshots (if relevant)

Does this MR meet the acceptance criteria?

What are the relevant issue numbers?

Closes #3453 (closed)

Edited by Nick Thomas

Merge request reports