A different flavor of the distributed transaction
summary
Argues that sagas replace two‑phase commit for long running distributed work and shows a working MicroProfile spec called MicroProfile LRA with a Narayana implementation (Narayana) and live Quarkus demos. Demonstrates annotations (start/mandatory/compensate/complete), HTTP header propagation of the LRA id, coordinator on localhost:8080 and services on 8081/8082, cancelOn/cancelOnFamily and end=false behavior, and that compensations run on failure so the demo actually executes and cleans up.
AI-written summary · about
chapters (16)
- 0:00Introduction
- 1:06ACID transactions explained
- 3:34Two-phase commit limits
- 7:51Saga pattern and compensations
- 14:49BASE model and eventual consistency
- 16:57Saga orchestration vs choreography
- 17:39MicroProfile LRA intro
- 18:22Demo with Narayana LRA
- 21:15Quarkus LRA service
- 24:06LRA annotations and compensations
- 28:42LRA ID mapping
- 30:08Extending saga across services
- 32:40Transaction propagation
- 35:11Error handling and LRA control
- 36:18MicroProfile LRA and resources
- 37:23Q&A: ACID vs BASE
related talks
50mTransactional solutions for microservices
46mAddressing the transaction challenge in a cloud-native world
55mTackling transactions in Microservices applications
01h 01mLost in transaction Strategies to manage consistency in distributed systems
30mThe Architecture of Microservice Transaction Based on Distributed Consensus with Kubernetes
30mData Consistency Patterns in Cloud Native Applications
54mTackling transactions in a Microservices application
52mArchitectural Patterns for Distributed and Decentralized Data