[Date Prev][Date Next] [Thread Prev][Thread Next] [Date Index] [Thread Index]

Re: improving documenation

On 4/28/21 9:53 AM, Thomas Lange wrote:
> Hi,
> I've created a proposal for updating the header seen in
> https://cloud.debian.org/cdimage/cloud/
> You can see it here
> https://public.cs.uni-koeln.de/lange/images-cloud-HEADER.html
> Any comments welcome.

Hi Thomas,

Thanks for this.

Is there a git for it? It's been a long time I want to add what I'm
describing below. Let me explain...

I would also very much welcome instructions on how to upload these
images. For example, at:

one can read how the OpenStack command line works. On top, we would need
to add a few explanation about the standard property. We should, IMO,
let our user do:

openstack image create \
        --container-format bare \
        --disk-format qcow2 \
        --property hw_disk_bus=scsi \
        --property hw_scsi_model=virtio-scsi \
	--property os_type=linux \
	--property os_distro=debian \
	--property os_version=11.0.0 \
        --file <FILE-NAME> \

The "--property hw_disk_bus=scsi --property hw_scsi_model=virtio-scsi"
bit is to instruct OpenStack to use the virtio-scsi driver rather than
the virtio-blk driver. The former is much nicer, exposing /dev/sdX
instead of /dev/vdX, and supporting FSTRIM, which is useful for
regaining free space (note: the virtio-blk driver also supports FSTRIM
in its latest version, but not all providers are running the latest Qemu).

OpenStack has standardized on some image properties, helping to describe
what the image is for. The documentation is here:

These are metadata only useful to our users, who can later on filter
using them. For example, this command:

openstack image list --property os_distro=debian

will then display only Debian. It'd be great if the majority of public
cloud providers could implement this.

Last, it'd be nice to instruct our users that the "generic" image (ie:
the one with all the drivers, as opposed to genericcloud) is a good fit
to use with "OpenStack baremetal", aka Ironic.

Reply to: