Multi-Region, Environment Isolated Ray Environments in the Notion Way
summary
Notion rearchitects AnyScale deployment to enforce true region-and-environment isolation by creating 8 clouds (one per region per environment), moving all jobs into dedicated clouds, and scoping permissions with scoped IAM roles. Infrastructure is codified via the Terraform provider (beta) to register clouds (two‑phase handshake), import legacy clouds, and standardize names for S3/KMS and secrets (region + dev/prod). Jobs migrate easily; services require parallel run and traffic gating for safe cutover.
AI-written summary · about
chapters (14)
- 0:00Introduction and team
- 1:52Notion and the Notion way
- 3:37Ray use cases at Notion
- 4:18Multi-region expansion challenges
- 6:04Infrastructure issues identified
- 8:09Redesign goals and principles
- 9:33Implementation approach
- 12:09Cloud and project isolation
- 13:12Step-by-step rollout process
- 15:16Adopting new tools and fixes
- 16:19Changes for engineers
- 17:20Next steps and future plans
- 18:02Lessons learned
- 19:15Q&A session
related talks
29mMulti Region Cloud Architecture for Data Residency Compliance
35mMulti Region Terraform Deployments
17mMulti Region Deployments with AWS CI/CD and Terraform
33mEnvironments as Cattle
37mThe Nth Region Project: An Open Retrospective
30mHow Sunrise Banks Built a Symmetric Multi Cloud Architecture with Terraform & Vault
39mTerraform Practices to Enable Infrastructure Scaling
41mAutomating Multi-Cloud, Multi-Region Vault for Teams and Landing Zones