36 Microservices Patterns & Anti-Patterns: The Definitive Reference (2026)

Q: What is the difference between microservices patterns and anti-patterns?

Patterns are proven architectural solutions to recurring problems in distributed systems — they represent best practices distilled from production experience. Anti-patterns are commonly adopted approaches that appear reasonable on the surface but lead to systemic failures, technical debt, or scalability bottlenecks in practice. Understanding both is essential: patterns tell you what to do, anti-patterns tell you what to avoid. This guide catalogues 26 patterns and 10 anti-patterns across six categories.

Q: How many microservices patterns should an architect know?

An architect working with microservices should have working knowledge of at least the core patterns in each category — API Gateway, Circuit Breaker, Saga Orchestration, Event-Driven Architecture, CQRS, and Sidecar at minimum. The full catalogue of 26 patterns covered in this index provides comprehensive coverage across infrastructure, resilience, data consistency, async communication, and AI governance. You do not need to implement all patterns simultaneously; start with the ones that solve your most pressing architectural challenges.

Q: What are the most critical microservices anti-patterns to avoid?

The three most damaging anti-patterns in production systems are Synchronous Blocking Calls (which cause cascading latency failures under load), No Idempotency on Critical Operations (which leads to double charges and data corruption during retries), and Tightly Coupled Services (which eliminates the independent failure domains that justify microservices in the first place). In AI systems, Unclassified Query Handling and Generic Single-Prompt Systems are equally damaging, leading to inconsistent safety and poor user outcomes.

Q: Why does this guide include AI and LLM governance patterns alongside classical microservices patterns?

In 2026, AI components — LLM inference services, classification pipelines, agent orchestrators, and guardrail engines — are deployed as microservices within larger distributed systems. They face the same challenges as classical microservices (resilience, scaling, observability) plus unique challenges around query classification, prompt management, safety governance, and confidence calibration. The AI governance patterns in this guide (Classification-First Architecture, Prompt Routing, Versioned Prompt Templates, and others) address these AI-specific architectural challenges using microservices principles.

Q: What is the Saga Orchestration pattern and when should I use it?

Saga Orchestration coordinates distributed transactions across multiple microservices using a sequence of local transactions with compensating actions. For example, an e-commerce checkout might reserve inventory, charge the payment, and create the order as separate steps. If the payment fails, the saga executes a compensating action to release the reserved inventory. Use it whenever a business process spans multiple services and requires eventual consistency — the most common scenarios are checkout flows, booking systems, and multi-step fulfilment pipelines.

Q: How does the Circuit Breaker pattern prevent cascading failures?

The Circuit Breaker monitors calls to downstream services and tracks failure rates. When failures exceed a threshold, the breaker opens and immediately returns a fallback response instead of attempting the call — preventing the calling service from waiting for timeouts that block threads and exhaust connection pools. After a cooling period, the breaker enters a half-open state and allows test requests through to check if the downstream service has recovered. This pattern is non-negotiable for any service with external dependencies, as it prevents a single failing dependency from taking down the entire system.

Q: What is the Outbox pattern and why does it matter for event-driven microservices?

The Outbox pattern solves the dual-write problem: when a service needs to update its database and publish an event, these two operations cannot be made atomic without special handling. The service writes the event to an outbox table within the same database transaction as the data change. A separate process (change data capture or polling) reads the outbox table and publishes events to the message broker. This guarantees that database changes and event publications stay synchronised, preventing the data inconsistencies that plague event-driven systems without this pattern.

Q: How do I choose between Sidecar and Ambassador patterns?

The Sidecar pattern handles cross-cutting concerns for inbound traffic to a service — mTLS termination, request logging, health checks, and telemetry. The Ambassador pattern is a specialised sidecar focused on outbound traffic — managing connection pooling, retries, timeouts, circuit breaking, and protocol translation for calls the service makes to other services. In practice, many service mesh implementations (Envoy, Linkerd) combine both roles in a single proxy. Use Sidecar when you need to abstract inbound concerns; use Ambassador when outbound call management is the primary challenge.

Q: What is Classification-First Architecture in AI systems?

Classification-First Architecture determines query intent and risk level as the foundational step before any model processing. Every incoming query is classified by category, complexity, and risk, and this classification drives all downstream decisions — which prompt template to use, which model to route to, what guardrails to apply, and whether to escalate to a human. Without classification first, AI systems treat all queries equally, leading to the Generic Single-Prompt and Monolithic Safety Rules anti-patterns.

Q: Will each pattern get its own dedicated deep-dive article?

Yes. This master index is a living hub page. Each of the 26 patterns and 10 anti-patterns will receive a dedicated deep-dive article covering the problem context, architectural diagrams, implementation guidance, when to use and when to avoid the pattern, and real-world examples. Articles are published in priority batches starting with patterns that have no existing coverage. As each article is published, the corresponding entry in this index will be updated with a direct link to the deep-dive.