[gate] nova-compute failing with python 3.6 + libvirt-python 6.8.0

Lee Yarwood lyarwood at redhat.com
Wed Oct 28 09:52:40 UTC 2020


On 27-10-20 09:08:17, Lee Yarwood wrote:
> On 26-10-20 13:32:20, Kashyap Chamarthy wrote:
> > On Mon, Oct 26, 2020 at 01:17:17PM +0100, Kashyap Chamarthy wrote:
> > 
> > [...]
> > 
> > > Great sleuthing so far.  (Not to mention the terribly cryptic orignal
> > > error message)
> > > 
> > > As a small addendum, was talking to Dan Berrangé from libvirt and
> > > pointed to the openAuth() method ... and  he wrote a quick
> > > 'libvirt-python' unit test for the openAuth() method:
> > > 
> > >     https://gitlab.com/libvirt/libvirt-python/-/merge_requests/28/diffs
> > > 
> > > ... and it seems to be working as expected (based on the CI result):
> > > 
> > >     https://gitlab.com/libvirt/libvirt-python/-/merge_requests/28/pipelines
> > > 
> > > ... it seems to validate several operating system platforms (and Python
> > > versions), including Ubuntu 18.04:
> > > 
> > >     https://gitlab.com/berrange/libvirt-python/-/pipelines/207610554
> > 
> > So ... Dan was able to reproduce the problem; the culprit was the the
> > combination of thread pool, Python 3.6.x and type hinting; the above
> > unit test from Dan didn't make use of 'tpool'.  He says:
> > 
> >     tpool + Python 3.6.x + type hinting is fubar
> > 
> >     If I add 'tpool' into my new unit tests, then it crashes & burns
> >     with the error msg Mel got 
> > 
> >     
> > [...]
> 
> I've pushed https://review.opendev.org/#/c/759831/ based on the feedback
> from danpb, lets see if it resolves the issue.

Thus far we've hit the following issues in the gate:

nova-grenade-multinode
----------------------

libvirt.libvirtError: internal error: missing block job data for disk
'vda'
https://bugs.launchpad.net/nova/+bug/1901739

nova-next
---------

volume delete fails because cinder-rootwrap lvs fails with exit code 139
Edit
https://bugs.launchpad.net/cinder/+bug/1901783

grenade
-------

60_nova/resources.sh:106:ping_check_public fails intermittently
https://bugs.launchpad.net/neutron/+bug/1463631

The nova-ceph-multistore job has also failed, timing out after g-api
issues that I've not been able to nail down.

frickler has suggested force merging this so we can open the gate back
up for other projects, I'm not against this by the ultimate decision
rests with gibi as PTL.

I've also just noticed that the change is now somehow in the check and
gate queues at the same time?! It has already passed the check queue a
few times so I'm not worried about this but I wonder if this is a zuul
bug of some kind?

-- 
Lee Yarwood                 A5D1 9385 88CB 7E5F BE64  6618 BCA6 6E33 F672 2D76
-------------- next part --------------
A non-text attachment was scrubbed...
Name: signature.asc
Type: application/pgp-signature
Size: 488 bytes
Desc: not available
URL: <http://lists.openstack.org/pipermail/openstack-discuss/attachments/20201028/bd28a9c0/attachment-0001.sig>


More information about the openstack-discuss mailing list