virt-install: changing default --os-variant behavior

[Cole Robinson]

On 9/22/20 11:59 AM, Peter Crowther wrote:
> On Tue, 22 Sep 2020 at 16:48, Cole Robinson <crobinso at redhat.com
> <mailto:crobinso at redhat.com>> wrote:
> 
>     On 9/20/20 4:46 PM, Peter Crowther wrote:
>     > On Sun, 20 Sep 2020 at 21:10, Cole Robinson <crobinso at redhat.com
>     <mailto:crobinso at redhat.com>
>     > <mailto:crobinso at redhat.com <mailto:crobinso at redhat.com>>> wrote:
>     > [...]
>     >
>     >     2) Default to --os-variant detect=on,name=<virtio-something>.
>     'give me
>     >     virtio' is representative of what most virt-install users
>     want. But this
>     >     adds some new corner cases, ex if anyone is using virt-install
>     with
>     >     windows up until now they could get away without specifying a
>     >     --os-variant and things would generally work, but now if we
>     default to
>     >     virtio windows out of the box is not going to install. I kinda
>     doubt
>     >     many people are using virt-install with windows though.
>     >
>     >
>     > As feedback, this is the single largest use case in the main
>     > virtualisation cluster I manage.  CentOS hosts, 90% Windows 7(!)
>     and 8.1
>     > guests.  We have our virt-install scripted to add a floppy drive with
>     > autoinstall file, virtio drivers, and a few other bits and pieces
>     like a
>     > minimal puppet client install (surprisingly non-trivial in Windows 7
>     > gold), so we could live with such a change; but please don't
>     assume that
>     > if the hosts are Linux then the guests are also likely to be Linux.
>     >
> 
>     But presumably you are already specifying an --os-variant though? In
>     which case you would be fine even if the default changes.
> 
> 
> Yes, we are.
> 
> On the principle of least surprise, I would suggest "Option 3" of your
> two suggestions: virt-install warns but succeeds for such an install.  I
> don't think it's unreasonable to default to slower emulated hardware
> rather than virtio as long as the user gets a reasonably helpful notice
> that virt-install doesn't know what else to do so has gone for a slow
> fail-safe.
> 

We already do this but it's still problematic in my experience. Most
people aren't intentionally opting into old hardware, they are getting
it by accident or ignorance.

Is #1 palatable if you consider we hard error, but print a suggestion
that includes: 'if you want the old behavior, add --os-variant
detect=on,name=generic` to your command line'. Some people implicitly
depending on the old behavior will have to one time update their
scripts. New people writing scripts will add that change after their
first attempt. The only problem I see is if people are running the same
script across drastically different versions of virt-install.

Actually now that I think about it, that's not a problem either.
`--os-variant auto` has been available for a decade, which explicitly
opts in to the current default behavior. So users will have a cross
version way to get the old behavior back.

Thanks,
Cole