Building Microservices: Designing Fine-Grained Systems, 2nd Edition

go site Preface Who Should Read This Book Why I Wrote This Book What’s Changed Since the First Edition? Navigating This Book Part I, Foundation Part II, Implementation Part III, People Conventions Used in This Book O’Reilly Online Learning How to Contact Us Acknowledgments I. Foundation 1. What Are Microservices? Microservices at a Glance Key Concepts of Microservices Independent Deployability Modeled Around a Business Domain Owning Their Own State Size Flexibility Alignment of Architecture and Organization The Monolith The Single-Process Monolith The Modular Monolith The Distributed Monolith Monoliths and Delivery Contention Advantages of Monoliths Enabling Technology Log Aggregation and Distributed Tracing Containers and Kubernetes Streaming Public Cloud and Serverless Advantages of Microservices Technology Heterogeneity Robustness Scaling Ease of Deployment Organizational Alignment Composability Microservice Pain Points Developer Experience Technology Overload Cost Reporting Monitoring and Troubleshooting Security Testing Latency Data Consistency Should I Use Microservices? Whom They Might Not Work For Where They Work Well Summary 2. How to Model Microservices Introducing MusicCorp What Makes a Good Microservice Boundary? Information Hiding Cohesion Coupling The Interplay of Coupling and Cohesion Types of Coupling Domain Coupling Pass-Through Coupling Common Coupling Content Coupling Just Enough Domain-Driven Design Ubiquitous Language Aggregate Bounded Context Hidden models Shared models Mapping Aggregates and Bounded Contexts to Microservices Turtles all the way down Event Storming Logistics The process The Case for Domain-Driven Design for Microservices Alternatives to Business Domain Boundaries Volatility Data Technology Organizational Mixing Models and Exceptions Summary 3. Splitting the Monolith Have a Goal Incremental Migration The Monolith Is Rarely the Enemy The Dangers of Premature Decomposition What to Split First? Decomposition by Layer Code First Data First Useful Decompositional Patterns Strangler Fig Pattern Parallel Run Feature Toggle Data Decomposition Concerns Performance Data Integrity Transactions Tooling Reporting Database Summary 4. Microservice Communication Styles From In-Process to Inter-Process Performance Changing Interfaces Error Handling Technology for Inter-Process Communication: So Many Choices Styles of Microservice Communication Mix and Match Pattern: Synchronous Blocking Advantages Disadvantages Where to Use It Pattern: Asynchronous Nonblocking Advantages Disadvantages Where to Use It Pattern: Communication Through Common Data Implementation Advantages Disadvantages Where to Use It Pattern: Request-Response Communication Implementation: Synchronous Versus Asynchronous Where to Use It Pattern: Event-Driven Communication Implementation What’s in an Event? Just an ID Fully detailed events Where to Use It Proceed with Caution Summary II. Implementation 5. Implementing Microservice Communication Looking for the Ideal Technology Make Backward Compatibility Easy Make Your Interface Explicit Keep Your APIs Technology Agnostic Make Your Service Simple for Consumers Hide Internal Implementation Detail Technology Choices Remote Procedure Calls Challenges Technology coupling Local calls are not like remote calls Brittleness Where to use it REST REST and HTTP Hypermedia as the engine of application state Challenges Where to use it GraphQL Challenges Where to use it Message Brokers Topics and queues Guaranteed delivery Trust Other characteristics Choices Kafka Serialization Formats Textual Formats Binary Formats Schemas Structural Versus Semantic Contract Breakages Should You Use Schemas? Handling Change Between Microservices Avoiding Breaking Changes Expansion Changes Tolerant Reader Right Technology Explicit Interface Catch Accidental Breaking Changes Early Managing Breaking Changes Lockstep Deployment Coexist Incompatible Microservice Versions Emulate the Old Interface Which Approach Do I Prefer? The Social Contract Tracking Usage Extreme Measures DRY and the Perils of Code Reuse in a Microservice World Sharing Code via Libraries Client libraries Service Discovery Domain Name System (DNS) Dynamic Service Registries ZooKeeper Consul etcd and Kubernetes Rolling your own Don’t Forget the Humans! Service Meshes and API Gateways API Gateways Where to use them What to avoid Service Meshes How they work Aren’t service meshes smart pipes? Do you need one? What About Other Protocols? Documenting Services Explicit Schemas The Self-Describing System Summary 6. Workflow Database Transactions ACID Transactions Still ACID, but Lacking Atomicity? Distributed Transactions—Two-Phase Commits Distributed Transactions—Just Say No Sagas Saga Failure Modes Saga rollbacks Reordering workflow steps to reduce rollbacks Mixing fail-backward and fail-forward situations Implementing Sagas Orchestrated sagas Choreographed sagas Mixing styles Should I use choreography or orchestration (or a mix)? Sagas Versus Distributed Transactions Summary 7. Build A Brief Introduction to Continuous Integration Are You Really Doing CI? Branching Models Build Pipelines and Continuous Delivery Tooling Trade-Offs and Environments Artifact Creation Mapping Source Code and Builds to Microservices One Giant Repo, One Giant Build Pattern: One Repository per Microservice (aka Multirepo) Reusing code across repositories Working across multiple repositories Where to use this pattern Pattern: Monorepo Mapping to build Defining ownership Tooling How “mono” is mono? Where to use this pattern Which Approach Would I Use? Summary 8. Deployment From Logical to Physical Multiple Instances The Database Database deployment and scaling Environments Principles of Microservice Deployment Isolated Execution Focus on Automation Two case studies on the power of automation Infrastructure as Code (IAC) Zero-Downtime Deployment Desired State Management Prerequisites GitOps Deployment Options Physical Machines Virtual Machines Cost of virtualization Good for microservices? Containers Isolated, differently Not perfect Windows containers Docker Fitness for microservices Application Containers Platform as a Service (PaaS) Function as a Service (FaaS) Limitations Challenges Mapping to microservices Function per microservice Function per aggregate Get even more fine-grained The way forward Which Deployment Option Is Right for You? Kubernetes and Container Orchestration The Case for Container Orchestration A Simplified View of Kubernetes Concepts Multitenancy and Federation The Cloud Native Computing Federation Platforms and Portability Helm, Operators, and CRDs, Oh My! And Knative The Future Should You Use It? Progressive Delivery Separating Deployment from Release On to Progressive Delivery Feature Toggles Canary Release Parallel Run Summary 9. Testing Types of Tests Test Scope Unit Tests Service Tests End-to-End Tests Trade-Offs Implementing Service Tests Mocking or Stubbing A Smarter Stub Service Implementing (Those Tricky) End-to-End Tests Flaky and Brittle Tests Who Writes These End-to-End Tests? How Long Should End-to-End Tests Run? The Great Pile-Up The Metaversion Lack of Independent Testability Should You Avoid End-to-End Tests? Contract Tests and Consumer-Driven Contracts (CDCs) Pact Other options It’s about conversations The Final Word Developer Experience From Preproduction to In-Production Testing Types of In-Production Testing Making Testing in Production Safe Mean Time to Repair over Mean Time Between Failures? Cross-Functional Testing Performance Tests Robustness Tests Summary 10. From Monitoring to Observability Disruption, Panic, and Confusion Single Microservice, Single Server Single Microservice, Multiple Servers Multiple Services, Multiple Servers Observability Versus Monitoring The Pillars of Observability? Not So Fast Building Blocks for Observability Log Aggregation Common format Correlating log lines Timing Implementations Shortcomings Metrics Aggregation Low versus high cardinality Implementations Distributed Tracing How it works Implementing distributing tracing Are We Doing OK? Service-level agreement Service-level objectives Service-level indicators Error budgets Alerting Some problems are worse than others Alert fatigue Toward better alerting Semantic Monitoring Real user monitoring Testing in Production Synthetic transactions Implementing synthetic transactions A/B testing Canary release Parallel run Smoke tests Synthetic transactions Chaos engineering Standardization Selecting Tools Democratic Easy to Integrate Provide Context Real-Time Suitable for Your Scale The Expert in the Machine Getting Started Summary 11. Security Core Principles Principle of Least Privilege Defense in Depth Automation Build Security into the Delivery Process The Five Functions of Cybersecurity Identify Protect Detect Respond Recover Foundations of Application Security Credentials User credentials Secrets Rotation Revocation Limiting scope Patching Backups Rebuild Implicit Trust Versus Zero Trust Implicit Trust Zero Trust It’s a Spectrum Securing Data Data in Transit Server identity Client identity Visibility of data Manipulation of data Data at Rest Go with the well known Pick your targets Be frugal It’s all about the keys Encrypt backups Authentication and Authorization Service-to-Service Authentication Human Authentication Common Single Sign-On Implementations Single Sign-On Gateway Fine-Grained Authorization The Confused Deputy Problem Centralized, Upstream Authorization Decentralizing Authorization JSON Web Tokens Format Using tokens Challenges Summary 12. Resiliency What Is Resiliency? Robustness Rebound Graceful Extensibility Sustained Adaptability And Microservice Architecture Failure Is Everywhere How Much Is Too Much? Degrading Functionality Stability Patterns Time-Outs Retries Bulkheads Circuit Breakers Isolation Redundancy Middleware Idempotency Spreading Your Risk CAP Theorem Sacrificing Consistency Sacrificing Availability Sacrificing Partition Tolerance? AP or CP? It’s Not All or Nothing And the Real World Chaos Engineering Game Days Production Experiments From Robustness to Beyond Blame Summary 13. Scaling The Four Axes of Scaling Vertical Scaling Implementation Key benefits Limitations Horizontal Duplication Implementations Key benefits Limitations Data Partitioning Implementation Key benefits Limitations Functional Decomposition Implementation Key benefits Limitations Combining Models Start Small Caching For Performance For Scale For Robustness Where to Cache Client-side Server-side Request cache Invalidation Time to live (TTL) Conditional GETs Notification-based Write-through Write-behind The Golden Rule of Caching Freshness Versus Optimization Cache Poisoning: A Cautionary Tale Autoscaling Starting Again Summary III. People 14. User Interfaces Toward Digital Ownership Models Drivers for Dedicated Frontend Teams Toward Stream-Aligned Teams Sharing Specialists Ensuring Consistency Working Through Technical Challenges Pattern: Monolithic Frontend When to Use It Pattern: Micro Frontends Implementation When to Use It Pattern: Page-Based Decomposition Where to Use It Pattern: Widget-Based Decomposition Implementation Dependencies Communication between in-page widgets When to Use It Constraints Pattern: Central Aggregating Gateway Ownership Different Types of User Interfaces Multiple Concerns When to Use It Pattern: Backend for Frontend (BFF) How Many BFFs? Reuse and BFFs BFFs for Desktop Web and Beyond When to Use GraphQL A Hybrid Approach Summary 15. Organizational Structures Loosely Coupled Organizations Conway’s Law Evidence Loosely and tightly coupled organizations Windows Vista Netflix and Amazon Team Size Understanding Conway’s Law Small Teams, Large Organization On Autonomy Strong Versus Collective Ownership Strong Ownership How far does strong ownership go? Collective Ownership At a Team Level Versus an Organizational Level Balancing Models Enabling Teams Communities of Practice The Platform The platform team The paved road Shared Microservices Too Hard to Split Cross-Cutting Changes Delivery Bottlenecks Internal Open Source Role of the Core Committers Maturity Tooling Pluggable, Modular Microservices Changing ownership Run multiple variations External contribution through libraries Change Reviews Synchronous versus asynchronous code reviews Ensemble programming The Orphaned Service Case Study: realestate.com.au Geographical Distribution Conway’s Law in Reverse People Summary 16. The Evolutionary Architect What’s in a Name? What Is Software Architecture? Making Change Possible An Evolutionary Vision for the Architect Defining System Boundaries A Social Construct Habitability A Principled Approach Strategic Goals Principles Practices Combining Principles and Practices A Real-World Example Guiding an Evolutionary Architecture Architecture in a Stream-Aligned Organization Building a Team The Required Standard Monitoring Interfaces Architectural Safety Governance and the Paved Road Exemplars Tailored Microservice Template Caution warranted The Paved Road at Scale Technical Debt Exception Handling Summary Afterword: Bringing It All Together What Are Microservices? Moving to Microservices Communication Styles Workflow Build Deployment Testing Monitoring and Observability Security Resiliency Scaling User Interfaces Organization Architecture Further Reading Looking Forward Final Words Bibliography Glossary Index