Lost in transaction? Strategies to deal with (in-)consistency in distributed systems
summary
Claims that distributed two‑phase commits don’t scale and shows a working alternative: model business transactions as Saga pattern workflows, implemented with a lightweight embedded engine (Java/Spring demo) that returns 202 Accepted and performs stateful retry and compensations offline until success. Uses REST to Stripe, mentions Kafka/Cassandra for durability, highlights idempotency zucchini as critical, compares choreography vs orchestration, and demonstrates the engine persisting state and retrying until the workflow completes.
AI-written summary · about
chapters (12)
- 0:00Start
- 2:17Fallacies of Distributed Computing
- 5:17Starbucks Does Not Use Two-Phase Commit
- 9:16Idempotency
- 14:29Stateful Retry
- 29:08Choreographer Approach
- 32:38Orchestration
- 40:00Test Reports
- 41:08Workflow Engine Market
- 41:44Open Source Frameworks
- 43:11Recap
- 44:53Axon Framework
related talks
01h 01mLost in transaction Strategies to manage consistency in distributed systems
36mLost In Transaction
01h 02mConsistency and Agreements in Distributed Systems
50mTransactional solutions for microservices
17mSaga Pattern for Distributed Transactions
30mData Consistency Patterns in Cloud Native Applications
31mLost in Transaction?
55mTackling transactions in Microservices applications