Serverless Applications - kimschles/schlesinger-knowledge GitHub Wiki

Self-Healing Serverless Applications AKA How to build systems that are resistant to failure

Nate Taggart, CEO of Stackery Gluecon: May 16, 2018

Common Serverless Failures for Lambda-Based Architectures

• Runtime Errors
  • Uncaught exception
  • Timeout
  • Bad state
• Scaling Errors
  • Concurrency limits (how high you can scale)
  • Spawn limits (how fast you can scale)
  • Bottlenecking

The Alternative: Self-Healing Serverless Apps

3 Design Principles:

1. Plan for failure
   • Know the service limits during your planning
   • Use self-throttling
   • Consider alternatives to AWS lambda
2. Standardize
   • Introduce universal instrumentation
   • Collect event-centric diagnostics
   • Give everyone visibility
3. Fail Gracefully
   • Reroute and unblock failed traffic
   • Automate known solutions
   • Notify a human

Research

• DB connection pool
• API Gateway
• Python try: except and raise (try-catch loop)
• Kinesis Stream
• Decorator Pattern