100 Temporal Mistakes and How to Avoid Them
summary
Practical checklist of production mistakes and fixes for Temporal: use schedule to close timeout as the real bound and prefer unlimited retries with exponential backoff (use the Temporal simulator to validate), require heartbeats for long-running or cancelable activities, make activities idempotent (HTTP idempotency keys, exists checks, or split steps), set heartbeat/start-to-close appropriately, enforce change versions at workflow start and use a replay testing harness for CI, and respect hard limits (≈2MB per payload, ~10k events, ~10MB history, 10s workflow task). GitHub and slides linked.
AI-written summary · about
chapters (11)
- 0:00Introduction
- 1:27Activity retry policies
- 9:33Activity best practices
- 11:17Workflow runtime limits
- 13:24Workflow determinism
- 13:44Workflow versioning
- 16:16Replay safety testing
- 19:49Child workflows and cancellation
- 21:54Workflow timeouts and timers
- 23:20Workflow guidelines
- 25:07Acknowledgments
related talks
30mOrchestrate Your Business Workflows with Temporal
23mDesigning a Workflow Engine From First Principles
26mTemporal: A Better Way to Build Modern Applications
32mWhy idempotency matters during durable execution in modern software stacks
26mMicroservices Madness: Application And Infrastructure Automation Using Temporal
42mHow to build distributed systems in JavaScript
01h 38mBuilding Reliable Distributed Systems
01h 47mWorkshop: How to Use Temporal to Build an MVP