Reimplement zero downtime deployments for single node instances with Puma
Summary
Since we switch our setup to Puma, we can't do zero downtime deployment anymore with our single node setup. We're aware that we could switch back to Unicorn for a year or so, but in the end Unicorn will be removed from Gitlab - at least that's my understanding.
As we now need to schedule a downtime for Gitlab updates/configuration changes/... things get more complicated as we need to announce a downtime beforehand in the company.
Also: By dropping zero downtime deployments you made it a Premium feature because only "Gitlab Premium" is allowed/supports scaled out architectures.
Proposal
Please find a way to bring back zero downtime deployments for single nodes setups running Puma.