The Multi‑Workload Challenge: Predictable Failover
summary
Builds a declarative DR platform that executes a four‑stage DR procedure YAML (pre‑scale, promote DB, dial traffic, reset) and runs a state machine that posts stage events to Kafka + semaphore state machine for agents (database agent, capacity agent, in‑cluster pre‑scale add‑on) to pick up. Backend uses Go and Python, frontend React. Claims pre‑scaling from live primary traffic avoids cold capacity, health gates and reversible steps keep promotes safe, and rehearsals make failover predictable and faster.
AI-written summary · about
chapters (9)
- 0:00Introduction to Regional Failover
- 0:59Understanding the Problem
- 3:46Principles of Disaster Recovery
- 6:02Designing a Reliable Failover System
- 7:55Concrete Example: DR Procedure YAML
- 8:53High-Level Architecture Overview
- 10:49Ensuring Safety and Reliability
- 11:53Outcomes and Key Takeaways
- 13:29Conclusion and Contact Information
related talks
27mFailover Without Fear: A Practical Guide to Active-Active...
05mDisaster Recovery and Contingency Planning for Platforms
41mScaling Reliability Engineering with Tools
46mChaos Experiments - Datacenter Stress Testing
30mResilience strategies to level up your workloads
22mFailing Over without Falling Over
25mAutomating Disaster Recovery: The Ultimate Reliability Challenge
47mBuilding Resilience