Using CPU & Memory Hotplug for VM Vertical Scaling
summary
KubeVirt uses CPU and memory hotplug for vertical scaling on x86 Linux guests: changing vCPU from 2 to 4 triggers live migration to a target pod where libvirt hotplug adds CPUs, and increasing guest memory from 1Gi to 2Gi updates requested and current after online resize. Memory cannot be reduced below the boot size. API uses a cluster-level VM roll strategy (live vs staged); future work targets in‑place pod updates to avoid live migration and VMI subresources for scaling and metrics.
AI-written summary · about
chapters (5)
- 0:00Introduction
- 1:06CPU & memory hotplug
- 3:59API design challenges
- 7:10Future plans for hotplug
- 10:30Q&A
related talks
29mMutating the immutable: device hotplug in KubeVirt
32mA journey through supporting VMs with dedicated CPUs on Kubernetes
43mKubeVirt - Kubernetes, Virtualization and Your Future Data Center
31mKubeVirt: What If Kubernetes Orchestrated Your VMs?
19mKubevirt and the Cost of Containerizing VMs
43mMaking Running Virtual Machines in a Kubernetes Cluster a Mainstream Activity
36mIt’s All for the Users. More Durable, Secure, and Pluggable. KubeVirt v0.53
33mRestart-Free Vertical Scaling for Kubernetes Pods
