Kubespray: Driving Cost-Efficiency for AI on Kubernetes
Talkat KubeCon + CloudNativeCon Europe 2025 · on Thursday, 2025-04-03· at 16:00· Platinum Suite | Level 3 | Room 1-2·
summary
Kubespray manages cluster lifecycle across cloud and bare metal and is used to avoid vendor lock‑in while mixing providers for cost savings. CI routinely deploys 10–15k test clusters per month and runs 20–50 cluster deployments per PR, spinning up to 200–300 VMs for validation. Production users attach DGX and A100 hardware and deploy the NVIDIA GPU Operator plus TensorFlow/PyTorch stacks, enabling dynamic preemption, test tear‑down, and multi‑cloud workload placement.
AI-written summary · about
chapters (8)
- 0:00Kubespray introduction
- 1:51Flexibility and compatibility
- 3:16Cost efficiency and vendor lock-in
- 5:41Real-world multi-cloud savings
- 7:06Vestia Transun case study
- 9:36Release stability and support
- 10:39Community contributions
- 11:21Q&A on Ansible
related talks
40mBuilding Custom GPU Clusters at Scale: Using Kubespray To Create High-Performance AI Infrastructure
32mDeep Dive: Kubespray (a sig-cluster-lifecycle's project)
32mDeep Dive: Kubespray (a sig-cluster-lifecycle's project)
34mDeep Dive: Kubernetes/Kubespray (Cluster Lifecycle) SIG
36mKubespray Intro
44mSIG Cluster Lifecycle: What's New in Kubespray
04mLightning Talk: Kubespray CI with KubeVirt on Baremetal
27mHarnessing Kubernetes for Scalable AI/ML Workloads