[openstack-dev] [nova][hyperv] hyper-v CI issues

Matt Riedemann mriedem at linux.vnet.ibm.com
Wed Nov 16 15:17:52 UTC 2016


cfriesen was asking in IRC today why this libvirt-only driver change in 
nova kept failing the hyper-v CI:

https://review.openstack.org/#/c/346263

http://64.119.130.115/nova/346263/9/

It looks like there are a few issues:

1. The hyper-v CI doesn't appear to be testing cfriesen's change, it's 
testing this:

https://review.openstack.org/#/c/273504

Because this is in the logs:

"HEAD is now at 431e019 Merge commit 'refs/changes/04/273504/19' of 
ssh://review.openstack.org:29418/openstack/nova into HEAD"

2. I also noticed it's cherry picking a Tempest change:

+ git fetch git://git.openstack.org/openstack/tempest 
refs/changes/49/383049/8
 From git://git.openstack.org/openstack/tempest
  * branch            refs/changes/49/383049/8 -> FETCH_HEAD
+ git cherry-pick FETCH_HEAD
[master ec1bd9a] wait for port status to be ACTIVE

Which is: https://review.openstack.org/#/c/383049/

There is quite a bit of discussion in that patch, but essentially it's 
trying to workaround the fact that the hyper-v driver isn't waiting for 
vif plugged events when using neutron, so the server goes ACTIVE before 
the networking is completely setup.

I think that all virt drivers in nova that are using neutron, which 
needs to be all of them now as nova-network is going away, should be 
implementing the vif plugging wait/timeout code - it was added 
specifically for the CI issues we used to have in the gate before the 
neutron jobs were voting.

So who is working on adding that support to the hyper-v driver? I think 
that basically trumps any feature work that the hyper-v team is trying 
to get done in Ocata.

-- 

Thanks,

Matt Riedemann




More information about the OpenStack-dev mailing list