On Fri, Nov 13, 2015 at 12:56 PM, Michael McGrath <mmcgrath(a)redhat.com> wrote:
On Fri, Nov 13, 2015 at 10:50 AM, Josh Boyer
<jwboyer(a)fedoraproject.org> wrote:
> On Fri, Nov 13, 2015 at 11:35 AM, Dusty Mabe <dusty(a)dustymabe.com> wrote:
>>
>>
>> On 11/13/2015 10:41 AM, Amanda Carter wrote:
>>>
>>> Hey folks, creating a separate thread for this longer term discussion.
>>> We're getting ready to release our first 2 week atomic update on Tuesday
and
>>> Dusty Mabe has raised 2 potential release blockers that were not part of
>>> automated testing. It's good that he caught them, but it's also a bit
of a
>>> stroke of luck. Since there is no official QE for this release, who should
>>> own verifying that there are no release blocking bugs prior to every
>>> automated release and escalating if there are? If no one raises the blocker,
>>> we'll have no way to block the release.
>>>
>>> This is something that we need an answer to fairly quickly since we
don't
>>> even have confidence that the current release is good other than the current
>>> reports. And we'll be taking this plunge again in just 2 weeks.
>>>
>>> Thanks for your attention to this,
>>>
>>
>> We should really have a blocker review process for these similar to the ones
>> we have for normal Fedora releases. Primary items that we should be
>
> Are these 2 week images built from updates that are currently in
> stable, or are they built from updates-testing as well?
>
> I ask because it matters for things outside of atomic. The release
> blocker review process works because it catches things _before_ they
> are released for general availability. If the 2 week atomic images
> are only composed from already stable updates, then the packages are
> already out there in the project otherwise.
>
> So if you have something "blocker" in an atomic image, you are now
> forced to wait for an update to make it through the entire fedora
> updates process before you can ship your 2 week image. For something
> like the kernel, that might very well mean you don't ship a 2 week
> image because the fix is not in stable in sufficient time.
>
> The atomic images might be better served by doing tests on
> updates-testing packages that are included to ensure that blockers
> don't otherwise show up as a surprise. I would also recommend
> reaching out to the package owners for each important package in
> advance so that the atomic sig is aware of what is planned for updates
> and such.
>
> josh
I'd also point out that the nature of Atomic's rollback features make
it better suited to less testing in that they can always roll back.
The problem that I see comes if there's a bug that lasts through
several releases causing a fairly bad trust scenario w/ upgrades.
--
Mike McGrath | mmcgrath(a)redhat.com | (312) 660-3547
Atomic | Red Hat Chicago |
http://projectatomic.io/
One other thing I forgot to mention, Dennis requested some additional
QA for this to do some basic smoke tests. We're working on setting
something up that can be used for a basic smoke test until something
more permanent can be setup. I've cc'd Ari who's familiar.
Ari, can you give a quick update on what your team is thinking at this
point? Something automated would help greatly for at least one part
of the 2 week release (though bugs would still need to be handled).
--
Mike McGrath | mmcgrath(a)redhat.com | (312) 660-3547
Atomic | Red Hat Chicago |
http://projectatomic.io/