Scaling Spinnaker for multi-region multi-cloud SaaS deployments
summary
Salesforce runs ~1,800 provider accounts, ~100k pipelines and ~800k non‑prod executions, and faces 70+ minute clouddriver cache startups and 100+ minute new‑account ramps. Workloads use sharding cloud driver to pin account subsets (reducing three instances’ RAM from ~9GB to ~3GB), enable lazy validation and split clouddriver into ro/rw, and replace slow Spring property loading to load 1,000 accounts in under a minute instead of an hour. Planned additions include a bulk pipeline API, Redis tuning, and DR via global DNS.
AI-written summary · about
chapters (12)
- 0:00Introduction
- 2:05Spinnaker at Salesforce
- 5:20Fault Tolerance
- 6:08Edgar Deployment
- 6:58Challenges
- 12:20Existing features
- 13:13Cloud Driver
- 16:24Resource Utilization
- 17:34Restricting Instances
- 19:58Sharding Instances
- 21:18Loading Time
- 24:21Closing comments
related talks
27mSpinnaker at Salesforce: Enterprise-Level Multi-Substrate Continuous Delivery and Release Management
28mSalesforce Status Report – What Are We Up To With Spinnaker?
36mSpinnaker - Land of a 1000 Builds
45mEvolution of Spinnaker Internals & Operations
33mPanel Discussion: Spinnaker Success - CD as a Service using Spinnaker
16mManaging Accounts Across AWS Partitions
01h 04mOpening Keynote [Full]
46mSpinnaker at Scale for Snap's Multi Cloud Service Mesh