A skeptic's first contact with Kubernetes

The author reflects on their initial skepticism towards Kubernetes, stemming from a lack of direct experience and a perception of unnecessary complexity. After engaging with the platform, they aim to clarify its core concepts, focusing on its operational mechanics rather than just its components. Kubernetes enables the management of workloads through control loops, which continuously adjust resources to maintain a desired state. Key components include Pods, Nodes, Services, and Namespaces, with Services providing stable access and load balancing for Pods. The author highlights the importance of Controllers in managing workloads, emphasizing the benefits of using ReplicaSets and Deployments for scaling and updating applications. They also discuss storage management, contrasting ephemeral volumes with persistent volumes, and introduce StatefulSets for managing stateful applications. The author appreciates the design of Services and the Controller pattern but raises questions about the stability of clusters under constant change, the lack of support for dynamic scaling based on various metrics, and the modularity of storage and networking implementations. They express frustration with the complexity of Kubernetes' configuration language, particularly the reliance on text interpolation and manual formatting in tools like Helm. Overall, the author finds value in Kubernetes' architecture but remains critical of certain design choices and the learning curve associated with its use.