Open Source

10 items

kubernetescontainersorchestrationcloud-nativedevopsinfrastructuremicroservicesintermediate

Kubernetes: Container Orchestration Platform

The operating system for the cloud that runs 84% of organizations' containerized workloads

Go|108,000 stars|Updated January 2024|50 min read

View on GitHub

Summary

Kubernetes (K8s) is a container orchestration platform that automates deployment, scaling, and management of containerized applications. It abstracts away individual machines into a unified computing surface. You declare what you want (a Deployment with 3 replicas), and Kubernetes figures out how to make it happen. The control plane watches the cluster state and continuously reconciles actual state with desired state through a set of controllers.

Key Takeaways

Declarative Configuration

You describe the desired state (3 replicas of my app), not the steps to get there. Kubernetes controllers continuously reconcile actual state with desired state. If a pod dies, the controller creates a new one automatically.

Control Loop Architecture

Every component follows the same pattern: watch for changes, compare desired vs actual state, take action to reconcile. This simple pattern scales from managing pods to managing entire cloud infrastructure.

Pod as the Atomic Unit

A Pod is one or more containers that share network namespace and storage. Containers in a pod communicate via localhost and can share files. Pods are ephemeral - they can be killed and recreated anywhere in the cluster.

Kubernetes emerged from Google's internal system called Borg, which had been running containers at massive scale since 2003. When Docker made containers accessible to everyone in 2013, the industry needed a way to orchestrate them at scale.

The problem Kubernetes solves:

Without orchestration, running containers in production requires manual work: - Which server has capacity for this container? - What happens when a server dies? - How do containers find each other? - How do you roll out updates without downtime? - How do you scale from 3 to 300 instances?

Kubernetes automates all of this. You declare what you want, and it figures out how to make it happen.

Before and After Kubernetes

Core concepts:

Cluster: A set of machines (nodes) managed by Kubernetes
Node: A single machine (VM or physical) that runs containers
Pod: The smallest deployable unit - one or more containers
Deployment: Manages a set of identical pods with rolling updates
Service: Stable network endpoint for a set of pods
Namespace: Virtual cluster for resource isolation

Summary

Key Takeaways

Declarative Configuration

Control Loop Architecture

Pod as the Atomic Unit

Service Discovery Built-in

Services provide stable DNS names and IP addresses for a set of pods. When pods scale up or down, the Service automatically updates its endpoints. No external service registry needed.

Premium Content

Trade-offs

Aspect	Advantage	Disadvantage
Declarative Configuration	Desired state is version-controlled, reproducible, and self-documenting	Learning curve to understand YAML schemas and the reconciliation model
Abstraction Layer	Portable across cloud providers; same manifests work on AWS, GCP, Azure, on-prem	Adds complexity; debugging requires understanding both Kubernetes and underlying infrastructure
Self-Healing	Automatic restart of failed containers, rescheduling on node failures	Can mask underlying issues; restart loops may hide bugs instead of surfacing them
Extensibility (CRDs)	Can extend Kubernetes to manage any resource type with operators	CRD proliferation can make clusters hard to understand and maintain
Control Plane Overhead	Rich features: scheduling, networking, storage, RBAC, observability	Significant resource overhead; not suitable for small deployments or edge devices
Networking Model	Flat network with built-in service discovery and load balancing	Network policies can be complex; debugging network issues is challenging
etcd Dependency	Strong consistency guarantees for cluster state	etcd is a single point of failure; requires careful backup and HA setup
RBAC	Fine-grained access control for multi-tenant clusters	Complex to configure correctly; easy to be too permissive or too restrictive

Open Source

Redis: In-Memory Data Structure Store

Apache Kafka: Distributed Event Streaming Platform