Bringing Kubernetes into Salesforce [B]
summary
Salesforce runs Kubernetes on bare metal with ~100 nodes across 10 clusters and a custom controller that translates a Git-stored deployment manifest into Kubernetes artifacts, signs promoted Docker images (image signing) and enforces per-service security profiles by isolating pods to dedicated nodes. Deployments trigger from approved pull requests and reach production in 30 minutes to production. Current workloads are stateless caches, monitoring and feature-flag services, with GoCD pipelines, etcd for control plane state, and plans for persistent volumes, Ceph and stateful sets.
AI-written summary · about
chapters (11)
- 0:00Start
- 4:13Looking for a New Way
- 5:58A Simple Model
- 6:45We Chose Kubernetes
- 8:01From Decision to Delivery
- 8:57The Right Launch Partners Are Key
- 15:48Deployment Manifests
- 17:04Overall Flow
- 19:07Deployments Go In Parallel
- 21:24There's More To Do
- 22:47Persistence
related talks
32mA Peek Inside the Enterprise Cloud at Salesforce
29mHow Salesforce built an extensible PaaS using CRDs saving devs 4200 hours
27mOperationalizing Kubernetes Sidecars in Production at Salesforce
37mRunning Kubernetes in a Manufacturing Line
53mImplementing Kubernetes on with confidence
35mKubernetes journey
28mBuilding and Managing Kubernetes with Kubernetes
19mPains and gains of Kubernetes for Developers, story from experience