Systems playing nice: Building resilient integrations in an unpredictable world
summary
Parkster’s backend uses a transactional outbox written inside the parking service database so message writes and parking starts commit atomically, with workers delivering events to downstream services; retries rely on idempotency keys and exponential backoff with jitter, failing fast via short timeouts, circuit breakers and per-customer bulkheads, and moving stuck items to a dead letter queue for human fix. Real example: a partner polling every 2 minutes with a 1 minute timeout floods DB; monitoring outbox length acts as a canary.
AI-written summary · about
chapters (10)
- 0:00Introduction
- 3:43Planning for uncertainty
- 8:17Building resilient integrations
- 13:36Outbox pattern for persistence
- 18:57Timeouts and retries
- 23:54Retry queues and dead letters
- 26:45Circuit breakers and bulkheads
- 31:23Balancing safeguards
- 32:52Monitoring and feedback
- 36:47Learning from failures
related talks
52mMovin beyond request-reply: how smart APIs are different
51mPatterns for building resilient software systems
01h 01mBeyond Pub/Sub - Advanced Messaging Patterns
01h 03mBeyond Pub/Sub - Advanced Messaging Patterns
44m3 common pitfalls in microservice integration and how to avoid them
53mEffective Microservice Communication and Conversation Patterns
18mGuaranteed messages delivery:Challenges of transactional Outbox & Inbox patterns
49mManaging Resiliency in Microservices