Open Stack

I wanted to provide an update on where we're at with the series of nova 
changes to integrate with the new style 3.27 volume attachments API.

This is building a foundation between both projects for supporting 
multi-attach volumes in Nova in Queens.

The series has 3 major changes: 
https://review.openstack.org/#/q/status:open+project:openstack/nova+branch:master+topic:bp/cinder-new-attach-apis

a) Updating swap volume to use the new volume attachment flow. I've been 
working on this with Ildiko Vancsa. Integration tests are passing on 
this with the new style attachments enabled at the end of the series. 
This is should be ready to go but has not had review from other Nova 
cores, so is at risk.

b) Updating live migration to use the new volume attachment flow. Steve 
Noyes has been working on this. I've reviewed and helped with some 
testing (verified with CI that volume-backed live migration is passing 
against this change with the new flow enabled with libvirt). No other 
Nova cores have reviewed this yet so it's at risk. Also, even if we get 
that change in, we have to also update the Hyper-v and XenAPI drivers 
which also support live migration.

c) Updating the compute API code to start using the new attachments API 
when (1) all computes are upgraded to the latest version and (2) Cinder 
3.27 is available in the deployment. Those constraints are meant to 
support rolling upgrades before the new code flows are running. The 
change is passing Tempest CI but has not gotten much Nova core review 
yet. I've done some review but not a deep review on the latest revision.

Risk: High - this is high risk mainly because of the lack of another 
Nova core being involved in the changes. John Garbutt was the other core 
helping with this series but he has not been active since he was laid 
off from OSIC, which was back in May.

We also need to add some upgrade testing in the Grenade project so that 
we can be sure a volume attached in Ocata is properly detached in Pike. 
Steve Noyes was investigating this.

-- 

Thanks,

Matt