The Crossplane Experience
summary
Company adopts Crossplane and the Kubernetes reconciliation model to give developers API-driven infra but hits scaling pain: over 10,000 managed resources in one cluster and resource creation taking hours to days. Early wins include fast Helm-driven provisioning and polymorphic CRDs via the Upbound AWS provider, but controller bottlenecks and AWS rate limiting drive a pivot. Team builds a custom Kubebuilder Terraform operator that spins ephemeral Terraform pods per TF object (50 microservices → 50 pods, 1,000 pods possible) and it runs problem-free for over a year.
AI-written summary · about
chapters (9)
- 0:00Introduction and background
- 3:14What is Crossplane
- 8:56How Crossplane works
- 11:02Early challenges with Crossplane
- 14:13Scaling issues with Crossplane
- 18:48Building a Terraform controller
- 22:22Terraform controller in action
- 24:53Why we pivoted away
- 25:54Final thoughts and Q&A
related talks
25mCrossplane: Build and Manage Infrastructure using Kubernetes
37mPlatformOps with Crossplane
30mGetting to Know Crossplane
31mLearnings From Providing A Platform API With Kubernetes And Crossplane
35mBuilding and Enabling APIs with Crossplane 1
33mEliminating the Gap Between Development and Production with Kuber...
10mLightning Talk: Notes from Crossplane Newbs - Jeremy Tanner & Gianluca Arbezzano, Equinix
17mCrossplane: Deploy Your Cloud Components from Your K8S Cluster