On 5/10/2016 4:51 PM, Matt Riedemann wrote:
For the libvirt imagebackend refactor that mdbooth is working on, I have
a POC devstack-gate change which runs with the lvm imagebackend in the
libvirt driver [1].
The test results are mostly happy except for anything related to migrate
(including resize to same host) [2][3].
That's because we're not testing with boot from volume [4].
This is a weird capability wrinkle that is not clear from the API,
you'll only find out that you can't migrate/resize on this host that's
using lvm when it fails. We can't even disable this in tempest really
since there isn't a flag for
'only-supports-resize-for-volume-backed-instances'. So this job would
just have to disable any tests that have anything to do with
resize/migrate, which kind of sucks since that's what we wanted to test
going into the libvirt imagebackend refactor.
Anyway, I'm dumping this before leaving for the day, maybe others have
some ideas here.
[1] https://review.openstack.org/#/c/314744/
[2]
http://logs.openstack.org/44/314744/3/check/gate-tempest-dsvm-neutron-full/57a083b/logs/testr_results.html.gz
[3]
http://logs.openstack.org/44/314744/3/check/gate-tempest-dsvm-neutron-full/57a083b/logs/screen-n-cpu.txt.gz?level=TRACE#_2016-05-10_20_36_07_446
[4]
https://github.com/openstack/nova/blob/00eccf56d01f4945ab46f246ab4fe751375b39be/nova/virt/libvirt/driver.py#L6960
Just to give an update on this. The tempest change to add a
volume-backed resize test is merged. It turns out we had a bug with our
Rbd image backend in libvirt that the test exposed, but that's fixed now.
I've got a change up to add an LVM image backend job to Nova's
experimental queue [1]. It has a dependency on a devstack change to
ignore clearing volumes during the job run and a dependency on a Nova
change for our blacklist regex of tests that won't pass - the
non-volume-backed resize tests.
Once we get this job in we can run it on-demand on the libvirt storage
refactor series, or any other change in the libvirt driver code that
touches the LVM paths.
[1] https://review.openstack.org/#/c/316298/
--
Thanks,
Matt Riedemann
__________________________________________________________________________
OpenStack Development Mailing List (not for usage questions)
Unsubscribe: [email protected]?subject:unsubscribe
http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev
