Clarification on migrating from Sidekiq queue selectors to routing rules for HA.
This documentation describes how to migrate from queue selectors to routing rules. However it seems to only consider the case where GitLab is a single instance (including all the services like Sidekiq and Puma).
Can you clarify what the procedure is for a 3K user reference architecture with multiple nodes and service tiers? For example, we have a set of dedicated Sidekiq nodes, a set of dedicated Puma nodes, etc.
We also need to migrate this in a zero-downtime fashion.
Ticket: https://gitlab.zendesk.com/agent/tickets/460927 (Internal Only)
Edited by Amaury Yacksmith