How to perform a multi-rails-node GitLab upgrade WITH downtime
Summary
We've had a few tickets and issues raised because:
- zero-downtime upgrades were performed on environments
- multiple minor releases were applied in one step
13.x GitLab included at least one set of migrations which had to be run strictly in sequence and so caught out customers who attempted this.
example: gitlab#335026 (closed)
We appear to only document multi-node upgrades using the zero downtime procedure
What's the procedure for customers running multiple Rails nodes (eg: the 2K ref arch) and [a] are happy with downtime and [b] want the option of applying multiple minor releases.
Assumption: our specified upgrade path is being followed.
Steps to reproduce
Perform a zero downtime upgrade from, say, 13.2 to 13.12
What is the current bug behavior?
There's no with-downtime upgrade procedure for scaled out environments.
What is the expected correct behavior?
A clearly signposted alternative to zero-downtime upgrades for customers who want to apply multiple releases.