Learning to Fall

Discover how to build resilient microservices in this 39-minute Devoxx conference talk by Rafael Ordog. Learn from the journey of transforming an unreliable monolith into a robust set of microservices, focusing on graceful failure handling when dependencies are down. Explore bad practices, motivations for good practices, and useful patterns for creating resilient systems. Delve into topics such as distributed transactions, the saga pattern, and saga execution coordinators. Gain insights on achieving atomicity and availability, implementing compensating actions, and designing idempotent operations. Walk away with practical knowledge on how to think about error cases and build more robust microservice architectures.

Intro
Remember the monolith?
Micro services
The journey
Car rental co.
Travel agency
Travel packages
Separate databases Multiple services
Fire and forget
Manual failure recovery
Let's automate recovery
Brute force
Distributed transactions?
2-phase commit
Scalability issues
Single point of failure
Enlightenment
What if we didn't stick to ACID?
The saga pattern
Original purpose
Definition: Saga
Sequence of actions
Idempotent compensating action
Definition: Idempotent
Happy path
The Saga guarantee
Atomicity vs Availability
Digging deeper
SEC Saga Execution Coordinator
Saga log
When to undo?
Forward recovery
Periodically retry
Each action needs a unique ID
Conclusion
Saga helps you think of the error case
Think of compensating actions
Think of idempotent actions