The Enterprise´s new Shoes
summary
Adidas moves core storefront to Kubernetes on Giant Swarm, running ~30 clusters across five regions on AWS with CI/CD and Prometheus monitoring; automated cluster provisioning lets teams recover a destroyed cluster in 20 minutes recovery. Load testing mistake exposes a 10x error — 300k concurrent test instead of 30k — and production sale traffic is ~99% bot traffic, causing ingress CPU/memory exhaustion. Fixes: increase connection-tracking, set ingress resource requests/limits, add RBAC/PSP and on-prem storage plugins; result: site survives subsequent sales and scales with pre-warmed clusters.
AI-written summary · about
chapters (12)
- 0:00Introduction and speakers
- 1:26From legacy to Kubernetes journey
- 7:48Kubernetes 1.0 release discovery
- 10:20Enterprise culture shift
- 15:20Introducing Giant Swarm
- 16:43Playground clusters and chaos
- 18:53Security and compliance in Kubernetes
- 22:24First production workload
- 26:47DDoS attack during shoe sale
- 31:03Cluster organization strategies
- 33:56Cyber week and future plans
- 35:08Q&A session
related talks
51mCompetitive retail architecture with microservices
25mTackling operational and capacity utilization challenges at a large scale enterprise deployment
56mMastering Cloud Scalability and System Designs Deep dive
23mCapability Diamond
45mEvolution of Spinnaker Internals & Operations
01h 05mSpeed & Scale: How to Get There
19mBuilding a resilient system with Frappe and Kubernetes
42mCatastrophes and Atrophies