Chaos Engineering for SQL Server
summary
Shows concrete experiments: uses PowerShell and T‑SQL to stop-service on a primary to force an AG failover, discovers default clustered role failover threshold (1 failure per 6 hours) breaks real-world failback and finds ~15–20 servers with that setting, changes it to 2 failures per hour and retests successfully. Shares a backup horror story where an instant restore stalls at 99% for five hours then fails, requiring a six‑hour manual restore. Demonstrates Kubernetes with 10 SQL Server pods and the Cube Invaders GUI stress tests that mostly recover but reveal capacity gaps.
AI-written summary · about
chapters (9)
- 0:00Introduction to chaos engineering
- 1:47Defining chaos engineering
- 4:36Real-world implementations
- 7:47Environment analysis
- 11:18Identifying failure scenarios
- 15:21Potential testing scenarios
- 24:20SQL Server failover experiment
- 35:17Kubernetes and SQL Server
- 42:12Conclusion and resources
related talks
46mChaos Experiments - Datacenter Stress Testing
32mImplementing Chaos Engineering in Kubernetes
01h 02mPractical Chaos Engineering breaking things on purpose to make them more resilient against failure
39mChaos: Breaking your Systems to Make them Unbreakable
31mPlanning for Failure using Chaos Engineering
01h 01mEmbrace chaos to achieve stability
42mChaos Engineering: Why Breaking Things Should be Practiced
51mDesigning for Failure: Build Resilient Services