Applying the Principles of Chaos to Serverless
summary
Applies chaos engineering to AWS Lambda by instrumenting clients and configs rather than killing servers, using X‑Ray and SSM Parameter Store to injectableAll function and an HTTP wrapper that shows 50% latency between 100ms–5s (X‑Ray traces show added 3.4s in an invocation). Proposes dynamic timeouts via context.getRemainingTimeInMillis, simulates DynamoDB throttling by lowering function concurrency, records timeout incidents as custom metrics, keeps experiments in staging first, and stops when key percentiles (95/99) or error rates shift.
AI-written summary · about
chapters (9)
- 0:00Start
- 1:03Practices of Chaos Engineering
- 4:40Step One
- 8:21Keeping those Experiments under Control
- 9:44Containment and Blast Radius
- 15:52Failure Modes
- 17:16Latency
- 35:48Failures Are Inevitable
- 37:37How Feasible Is It To Run Chaos Engineering on a Deployment Pipeline
related talks
44mApplying principles of chaos engineering to serverless
34mChaos engineering for serverless
49mPerforming Chaos in a Serverless World
24mBreaking Serverless Things on Purpose: Chaos Engineering in Stateless Environments
23mSimplifying design complexities using Serverless Architecture on AWS
42mNo Server Is Better Than No Server: Building Blocks for Serverless Architectures
25mWhere Did All My Servers Go?
44mServerless Architecture on Production: Good, Bad and Ugly Parts