--- dts/vm_images/README.md | 64 +++++++++++++++++++++++++++++++++++++++++ 1 file changed, 64 insertions(+) create mode 100644 dts/vm_images/README.md
diff --git a/dts/vm_images/README.md b/dts/vm_images/README.md new file mode 100644 index 0000000000..1a59810d43 --- /dev/null +++ b/dts/vm_images/README.md @@ -0,0 +1,64 @@ +# DTS VM Images + +This folder contains utilities to create VM +images for use in virtio testing. + +## Host Requirements + +The host MUST have qemu/kvm with libvirtd installed +and set up. + +The host MUST be the same architecture as the VM +you are building. + +The host MUST have podman and either docker or have podman +aliased as docker (running "docker" calls podman). + +## Creating a VM + +Use the "create_vm_image.py" script to create the vm image. +If you do not have the required containers on your system, +it will build them. + +The root password it asks for is what to set the VM's +root password to, not the root password of the system +you run the script on. + +``` --debug ``` will enable debug output from guestfs +tools. This produces a lot of output and you shouldn't +use it unless something is going wrong. + +The base image MUST be a "cloud ready" or "prebuilt" +image, meaning you cannot use an installer ISO. It also +must be in the qcow2 format, (use qemu-img to convert it). +Most distros will have a "cloud image" which is in the +correct format. This base image will not be modified +by the build script. + +The output image is where all of the modifications go and +it is the image that you should use with DTS. + +## Supported Distros + +Currently, only RHEL 8 family distros and Ubuntu 20.04 are +supported. Debian might work, but it is untested. Most +testing has gone to Ubuntu 20.04. + +## Architectures + +Due to the way that guestfs tools work, they must run +under kvm, but the host needs to have a kernel image +that can be used to boot the VM. It may be possible +to work around this issue using containers, but +several days of experimentation kept running into +more and more complex issues with the interactions +between libguestfs and docker/podman. As such, +your best bet is to build your VMs on either a +bare-metal system of your desired architecture +or inside a VM already being emulated as your desired +architecture. This second approach may run into +issues with the hypervisor, since not all hypervisors +support nested virtualization by default. Since you need +an appropriate kernel image installed as well, it may +be easiest to build VMs using whatever distro you already +use for most of your servers. -- 2.20.1
