Maintainable Python Deployments at Scale: Decoupling Build from Runtime

摘要

Modern Python deployments often default to bundling application code, dependencies, and system configuration into Docker-based deployments. While this approach provides portability and fast startup, it can grow complex over time—especially in organizations with many teams and shared infrastructure—leading to maintainability and scalability challenges. This talk explores how Python-native packaging and environment management can support lightweight, production-ready deployments without sacrificing Docker's benefits. We’ll look at how tools like zipapp (PEP 441) can bundle Python applications, and share strategies for caching Python dependencies within virtual environments to improve startup performance. We'll review examples of how these are integrated into real-world systems. Whether you’re deploying to EMR, Kubernetes, or other infrastructure, you’ll leave with practical patterns—and a clearer mental model for architecting and deploying Python applications at scale.

說明