In most resource managers, we can set the container's CPU usage and memory usage. But I'm curious about the technical reasons for not supporting disk I/O resource allocation for containers.

CodePudding user response：

According to Using cgroups to limit I/O this requires cgroups v2 and that is quite recent feature for container runtimes.

For Kubernetes support, you should probably follow initial kep for qos of storage v0.1 and cgroups v2 was added as alpha feature in Kubernetes 1.22.

So it looks like there is work in progress on this.