Introduction

This document describes how to deploy and enable the vCPU hotplug feature on a Kunpeng server running the openEuler OS.

This feature increases or decreases the number of vCPUs of a running VM without interrupting services.

Nowadays, OSs must run reliably for a long time and have a powerful fault tolerance mechanism, that is, Reliability, Availability and Serviceability (RAS). Availability requires that minor problems in the system do not affect the normal running of the entire system. In some cases, hot plugging can be performed to replace faulty components, ensuring that the system downtime is within a certain range. In virtualization scenarios, dynamic scheduling of resources is also required in addition to RAS. When creating a VM, the customer cannot accurately predict the pressure of services on device resources. If the pressure on a device in a VM stays too high or too low for a long time, the service running speed is restricted or resources are wasted. vCPU hotplug can effectively solve this problem.

Resource scaling is one of the major advantages of cloud computing, and vCPU hotplug is one of the key technologies to achieve scalable CPU computing power. vCPU hotplug provides the following benefits:

• Accelerated VM startup, particularly for lightweight secure containers such as Kata containers. In this scenario, a container is initially configured with a single vCPU, with additional vCPUs being hot added after the container is started.
• On-demand resource scalability to reduce service costs. Developers can increase or decrease the number of vCPUs online based on service load requirements.