On Wed, May 4, 2016 at 1:13 PM, Dusty Mabe <dusty(a)dustymabe.com> wrote:
On 05/04/2016 05:03 AM, Peter Robinson wrote:
>> On 05/03/2016 07:34 PM, Peter Robinson wrote:
>>> On Tue, May 3, 2016 at 10:27 PM, Dusty Mabe <dusty(a)dustymabe.com>
wrote:
>>>> If we don't do this it will try to use grub2 but will fail
>>>> because we removed the packages from the package set.
>>>
>>> And does extlinux work in all the Vagrant use cases (I've never used
>>> it)? My understanding of using extlinux was because it was small and
>>> works on AWS for those images, in the case of vagrant it's 40Gb in
>>> size so size is not really an issue here and grub2 likely works better
>>> in the developer workstations (what ever Windows/Mac user use for
>>> hypervisors) use case so it might be better to explicitly add the
>>> packages to the vagrant image rather than move it to extlinux.
>>>
>>
>> Hey Peter,
>>
>> Thanks for bringing this up. I'll check that it works on the various
>> targets.
>>
>> I had previously suggested that we go back to using grub2 (like we did
>> in F23) for fedora cloud base (both base and vagrant images) but
dgilmore/mattdm,
>> and I decided to stick with extlinux for the space savings. This assumes that
>> we don't hit any problems in testing.
>>
>> I'd like to keep the cloud base and the cloud base vagrant image using
>> the same bootloader for consistency. If we hit problems with extlinux
>> for vagrant then I'd prefer we switch both cloud base and cloud base
>> vagrant images back to grub2. I'd prefer to not have them differ on
>> that front.
>>
>> What do you think?
>
> Personally not bothered either way. I know a lot of hypervisors are
> now defaulting to uEFI for virtual BIOSes now because it's needed for
> recent versions Windows etc and I'm not sure extlinux works in that
> use case. So it's probably worth seeing where
> VMware/Parallels/VirtualBox/HyperV are headed in this regards and if
> it works there. We can now actually distribute a uEFI firmware with
> Fedora (edk2-ovmf package currently in f25, built for f24) so it might
> be worth even starting with that to see if extlinux boots with it.
>
> From a non x86 cloud image using grub2 would be preferable for me with
> my secondary arches hat on as it means we could end up using a single
> kickstart, but it's a minor point/advantage.
I just want to make sure I understand that last point correctly:
- For non-x86 arches it is better if we are using grub
Is that a correct statement?
Sort of, more correct is "for non x86 arches grub is the only option"
IE there's no extlinux for Power64 for example, and in the arm* case
the extlinux support is not really the same as it is on x86 (it's not
a bootloader per say there, but more the firmware supports the
extlinux config to boot linux directly).
Peter