Don’t clone infra, isolate data: ephemeral PR environments in a shared-env
summary
Practical case study of running ephemeral PR preview environments inside a shared Kubernetes cluster without cloning full infra for each PR. Covers concrete patterns for data isolation in 2025—schema-per-env for Postgres, RabbitMQ vhosts/headers, tenant headers between services, metadata in Stripe/Paddle webhooks, and propagating Temporal context and lifecycle in CI/CD—plus PHP snippets, pitfalls, and templates to run dozens of parallel previews while reducing cost and speeding CI.
AI-written summary · about
chapters (20)
- 0:00Introduction
- 3:13Trunk-based development
- 5:42Data isolation in large systems
- 8:31Ephemeral environments principles
- 11:45Context-bundle schema solution
- 15:01Request routing via service mesh
- 17:30PostgreSQL data isolation
- 20:04RabbitMQ data isolation
- 21:49Key-value store data isolation
- 22:54N2N tests and CI/CD
- 25:47Policies and SDLC for success
- 27:36Summary and results
- 29:00Q&A: database drop risks
- 29:45Q&A: isolation architecture
- 31:54Q&A: implementation timelines
- 33:40Q&A: approach comparison
- 35:28Q&A: team scaling
- 35:49Q&A: third-party metadata
- 37:31Q&A: fixtures and migrations
- 38:14Q&A: schema changes
related talks
25mImprove Developer Productivity with Ephemeral Environments
34mThe Full Stack Preview: Kubernetes, ArgoCD, and Ephemeral Databases
16mStaging Environments: A Blueprint for Pre-production
33mA Journey To Ephemeral Test Environments With EKS and ArgoCD
27mEphemeral Environment Practices - the Good, Bad, and Ugly
31mImplementing Preview Environments with GitOps in Kubernetes
36mDesigning Staging Environments
22mEphemeral Environments Paradigm Powered by Spin...