The original title is "What a fake production stack taught me about CI/CD"

Short summary

The author built a self-contained CI/CD sandbox using Jenkins, Ansible, and Docker containers simulating a real dev/staging/prod fleet over SSH. Key lessons include matching test environment fidelity to the failure modes you want to catch, using Docker-outside-of-Docker (DooD) over Docker-in-Docker for simplicity while understanding its security tradeoffs, and making infrastructure shortcuts explicit named decisions with stated blast radius rather than hidden implementation details.

• •Test environment fidelity should match the failure modes you want to catch — skipping SSH skips all SSH-shaped problems
• •DooD (Docker-outside-of-Docker) is simpler than DinD but grants root-equivalent host access via docker.sock
• •Infrastructure shortcuts must be named decisions with stated blast radius, not unexamined implementation details

Generated with AI, which can make mistakes.