Running an API Gateway on Istio - Pitfalls and Learnings
summary
Runs Intuit’s API Gateway on Istio at scale: front door for billions of requests and 1.1 million RPS across thousands of pods in 40 namespaces and 13 clusters. Migration uses Envoy sidecars, Istio, Admiral for multicluster config and Admiral partitioning to avoid config blast, Argo for rollouts and Argo workflows for region failover. Key fixes: added hybrid TLS on the sidecar, tuned graceful shutdown, implemented bin‑packing and canaries, and reworked autoscaling to avoid CPU throttling; outlier detection for sparse backends remains a work in progress.
AI-written summary · about
chapters (13)
- 0:00Introduction
- 1:20What is an API Gateway
- 3:17Why inhouse development
- 8:34API Gateway Architecture
- 10:44API Gateway Pitfalls
- 12:12Graceful Shutdown Pitfalls
- 13:34Configuration Overload Pitfalls
- 14:42Outlier Detection
- 15:46Autoscaling Resource Limits
- 17:59CPU Throttling
- 19:14Overview
- 19:35Questions
- 23:49How do you achieve multi tendency and isolation
related talks
25mThe Future of API Management with Everything as Code
26mWhere API management meets the service mesh
55mMastering GraphQL: Strategies from the Experts
24mLeveraging Gateway API for Kubernetes Native API Management
29mTaming the Traffic: Select...
27mEnterprise API management in agile integration
42mAPI Gateways - The Good, the Bad and the Ugly
26mAPI Gateway, A Success Story