[openstack-dev] [nova][qa] Testing with ephemeral/swap disks
Diana Clarke
diana.joan.clarke at gmail.com
Tue Nov 8 05:32:22 UTC 2016
Ah yes, this gem ;)
This was my first attempt at a Tempest patch; my first contribution to
that project ever.
It was met with:
"this is the wrong approach and can never land like this"
"This as a separate test is also quite frankly overkill and unecessary."
"But if you really must have this verified in tempest..."
"But as this is today it can't ever land."
And then silence.
It was a truly welcoming experience. Perhaps a template we can use for
all new contributors going forward ;)
That said, I am actually interested in contributing further on this
front. Extend an olive branch, and point me in the right direction.
--diana
On Mon, Nov 7, 2016 at 11:03 AM, Matt Riedemann
<mriedem at linux.vnet.ibm.com> wrote:
> Nova has a known gap in integration testing of ephemeral/swap disks,
> especially around resizes and migrations.
>
> In Newton, Diana Clarke worked on a Tempest patch to verify the size of
> disks before and after a resize:
>
> https://review.openstack.org/#/c/338411/
>
> The QA team took issue with that patch because it creates specific flavors
> which might not work in all clouds, which is a valid concern.
>
> At the summit some of the Nova people talked about what if we just added
> ephemeral/swap to the flavors that devstack creates and puts in tempest.conf
> - then we'd get some implicit coverage of those by default. That doesn't
> really help non-devstack CI systems, but I'm not entirely sure it matters,
> at least right now.
>
> If we did have a specific test for this, like Diana's above, then we could
> just skip the test if the flavors in tempest.conf don't have ephemeral/swap
> defined. That would allow us to test this in the upstream CI, and external
> CI could also test it if they wanted to, but it wouldn't break external CI.
>
> Would that be OK? Or are there other issues with that approach?
>
> Alternatively we could work on some in-tree functional tests if those would
> work, but someone would have to investigate that. Note that the existing
> functional tests in Nova are not running libvirt, or glance, or any other
> external service - they only run a wsgi app server and a database (sqlite by
> default).
>
> There has been talk of creating a nova-dsvm-integration job (there are
> differing opinions on if that should use devstack or not) which wouldn't run
> Tempest tests, just in-tree integration tests. So they could test things
> that aren't going through the REST API, like the image cache. The problem
> with this extra job is the lack of warm bodies to work on a POC.
>
> So having brought these up - does anyone want to volunteer to explore these
> options or have issues with either one?
>
> --
>
> Thanks,
>
> Matt Riedemann
>
>
> __________________________________________________________________________
> OpenStack Development Mailing List (not for usage questions)
> Unsubscribe: OpenStack-dev-request at lists.openstack.org?subject:unsubscribe
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
More information about the OpenStack-dev
mailing list