evanfoster / kata-virtiofs-naive-rw-benchmark

Test that simulates a very naive sequential R/W workload.

Geek Repo:Geek Repo

Github PK Tool:Github PK Tool

Before running your tests, ensure your KUBECONFIG is defined (if not using ~/.kube/config) and your context is set.

To run the tests, simply run make, or make run. If you want to specify a namespace other than default, run NAMESPACE=foo make

After running your test, you'll need to make clean. This isn't handled automatically because doing so would nuke the pod logs.

The image is built and pushed using make build. They go to my Docker Hub namespace by default, but this can be overridden with the IMAGE_NAME environment variable.

This test depends on the following:

  • Docker/containerd nodes as a control
  • CRI-O nodes with Kata
    • The nodes should be labeled and tainted with node.kubernetes.io/kata-containers=true
    • Kata should be configured to use virtio-fs
    • Two runtimeClassNames must exist:
      • kata-qemu, for Kata
      • runc, for standard runc

About

Test that simulates a very naive sequential R/W workload.


Languages

Language:Python 66.0%Language:Shell 23.0%Language:Makefile 8.4%Language:Dockerfile 2.7%