Enabling Infrastructure as Code at LinkedIn
summary
LinkedIn builds a self‑hosted Terraform-as-a-service that bootstraps fabrics (manage and unmanaged modes), handles state, automatic Terraform version upgrades, and scales horizontally via API servers and worker queues. Key features: one-click import of existing resources into generated IaC, per‑user sandbox workspaces with auto‑destroy, and drift remediation that auto‑creates PRs from state diffs. Results: six functional fabrics in two years, fabric bootstrap time drops from 6–9 months to about one month, ~1,950 workspaces, 8,000 avg workflows/month (36,000 peak).
AI-written summary · about
chapters (13)
- 0:00Introduction and LinkedIn overview
- 2:13Need for Infrastructure as Code
- 5:03Terraform as a Service goals
- 7:56Moving to declarative workflows
- 9:45TFaS architecture and MVP
- 10:31One Click Import feature
- 13:39Sandbox workspaces for testing
- 16:04Developer experience enhancements
- 17:56Drift detection and protection
- 20:48Drift remediation workflow
- 25:20Progress and adoption metrics
- 27:09Why not Terraform Enterprise
- 29:14Conclusion and acknowledgments
related talks
22mEnabling Infrastructure as Code
52mInfrastructure as Code with Terraform
27mAdvanced Multipurpose Infrastructure: A Self-Managed Cloud Environment with HashiCorp Terraform
27mCase Study: Re-Thinking Our Infrastructure Tooling
55mSelf-service infrastructure is no longer a dream
25mRebuilding Infrastructure (as Code)
22mTerraform Hybrid: Unifying Your Physical and Cloud Datacenters
21mAccelerating cloud transformation with Terraform: A World Bank success story