Zero-Downtime Postgres Engine Upgrades

[jimlerza]

As a DevOps engineer, so that I can keep DAWSON's postgres engine up to date without causing frequent scheduled downtime, I need to be able to perform postgres engine upgrades with zero downtime.

Pre-Conditions

• None

Acceptance Criteria

• Utilizes Aurora RDS' native blue/green deployments for the primary cluster
  • Downtime is acceptable for the replica if it makes implementation easier
• No data loss during switchover (writes attempted during the switch-over should be retried by the client or queued by the backend)

Pain Avoided/Frustration Saved

• Nobody likes scheduled downtime

Breadth/Pervasiveness of Problem

• We don't have a system for regular postgres upgrades, so we run the risk of falling significantly behind again

Complexity of the Problem (Low, Medium, High) and Why it's Complex

• High

Notes

• https://docs.aws.amazon.com/AmazonRDS/latest/AuroraUserGuide/blue-green-deployments-creating.html