[ovirt-devel] Out of the box experience on Fedora - was: Re: [ovirt-users] [QE][ACTION REQUIRED] oVirt 3.5.1 RC status - postponed

[Francesco Romani]

----- Original Message -----
> From: "Sandro Bonazzola" <sbonazzo at redhat.com>
> To: Users at ovirt.org, devel at ovirt.org
> Sent: Tuesday, November 25, 2014 7:07:14 AM
> Subject: [ovirt-users] [QE][ACTION REQUIRED] oVirt 3.5.1 RC status -	postponed

[...]
> The following bugs have been keyworded as Regression and not marked as
> blockers:
> 1165336	virt		ASSIGNED	FC20 qemu needs kvmclock bugfixes

Hi, 

I'd like to elaborate a bit more here for the sake of the openness.
TL;DR version

- no actual regression in oVirt, meaning 'we applied a patch and we broke something'.
- "fix" is underway - to propose the relevant patches to Fedora QEMU package

Long(ish) version

1 some time ago we adhered to QEMU/KVM clock settings recommendations (see
  https://bugzilla.redhat.com/show_bug.cgi?id=1053846)
2 these recommendations are *still* valid as today - I just checked with upstream developers
  while investigating bz1165336
3 these recommendations may have surprising effects like disabling HPET clock
4 on some old(ish) upstream QEMUs, disabling HPET may hurt migrations - hence bz1165336
5 only *very* recent QEMUs (2.2.0rc0!) have the fixes, which are about improving
  kvm clock, while HPET clock is still not recommended (see #2 above)
6 if the qemu-kvm-rhev is used (available in the oVirt repo), the experience is significantly better


However, the reporter *has* a very valid point, which motivated me to write me this mail:
a. F20 is a supported platform
b. I *guess* Fedora is the platform of choice to try out QEMU and to initially play with it
c. out-of-the box experience with oVirt and Fedora is cumbersome, many steps and tunings are needed.
   This may annoy users - without a valid reason!
d. hence there is an unneededly high first step to try out oVirt, and this is hurting the project

Now, I'd like to raise this question

* it is true that Fedora is the platform of choice to try out and evaluate oVirt?
* if so, is the experience on Fedora streamlined enough or could it made simple, hence we could
  have a better vector to spread oVirt?
* what we could do, as oVirt project, to improve the above?

Feedback welcome

-- 
Francesco Romani
RedHat Engineering Virtualization R & D
Phone: 8261328
IRC: fromani