Ep02 - Ask Me Anything about DevOps, Cloud, Kubernetes, Platform Engineering,... w/Scott Rosenberg

The main goal of the video is to provide practical, real-world guidance on Kubernetes best practices, particularly for enterprise environments, with a focus on architecture, tooling, and operational decisions. Scott emphasizes simplicity, standardization, and avoiding overcomplication when deploying Kubernetes.

Key Takeaways:
• Organize Kubernetes manifests per application in separate repositories, keeping source code and manifests separate to avoid cascading failures 2:05.
• Prefer pure YAML manifests in environment-specific repositories rather than Helm templates, to avoid complex templating and mental overhead 4:00.
• Avoid running monolithic or large applications in Kubernetes if they require direct hardware access or frequent manual updates—these are better suited to traditional deployment models 5:35.
• Use industry-standard tools (e.g., Prometheus, CNI from cloud providers) and avoid overloading clusters with tools—standardization improves observability and reduces troubleshooting complexity 25:50.
• For on-premises deployments, use vendor defaults (e.g., containerd in AWS, EKS) and avoid specialized runtimes like Kata or gVisor unless absolutely required due to overhead 46:06.
• Avoid using ECS on AWS; it has become overly complex and is less valuable than Kubernetes, which is now dominant in AWS and offers better ecosystem support 35:22.

Kubernetes is best used when it adds value—such as enabling dynamic scaling and robustness—but should not be over-applied to applications that don’t benefit from its features.

Sources:

• 2:05 Discussion on mani