[openstack-dev] [nova] review priorities ocata summit session recap

Matt Riedemann mriedem at linux.vnet.ibm.com
Wed Nov 16 21:57:18 UTC 2016

The final design summit session at the Ocata summit was, as usual, the 
review priorities session. The full etherpad is here:


Given the short cycle and restricted core reviewer bandwidth we're 
really only making resource providers and cells v2 review priorities for 
the release, along with some other things those depend on.

Resource Providers

There are three efforts here which can be worked concurrently, but the 
priority order is:

1. Making nova-scheduler use the placement service. Sylvain Bauza is 
working on this.

2. Handling shared storage/IP pools. Chris Dent is working on the server 
side aggregates changes for this and Jay Pipes is working on the client 
side pieces in the resource tracker.

3. Custom resource classes. Jay Pipes is leading the development on this 
and already has most, if not all, of the code up for review, some if it 
already merged.

Cells v2

1. Scheduler interaction. Dan Smith is leading the development work on 
this, lots of code is up for review and some of the series is already 

2. Adding the multi-cell simple python merge sort / filtering for 
listing instances. Dan Smith signed up to work on this.

3. Quotas. Melanie Witt is owning this, and as pointed out in an earlier 
recap on this specific session we might change how we do quotas for 
cells v2 in the API.

4. CI testing. I'm signed up to work on this. Some of the work has 
already gone into switching Ocata CI jobs (except cellsv1) to Neutron by 
default, making nova-network fail to start unless you're in a cells v1 
environment (dansmith wrote that patch), and then enabling cells v2 in 
master branch jobs by default (there are grenade changes needed for 
this). Eventually I'll be working on multinode (multicell) support here too.


There are two other peripheral efforts that are a review priority 
because of cells v2 being dependent on them.

1. Restricting the server list filters/sort parameters. Kevin Zheng and 
Alex Xu are working on this.

2. Moving neutron port creation to conductor. This is a dependency for 
supporting routed networks which is a dependency for using neutron with 
multiple cells. John Garbutt already started working on this in Newton 
so this is just a continuation of that work in Ocata.


There are a few non-priority but still notable mentions of things we're 
going to keep working on throughout the release:

1. Getting an agreement on the Cinder API rework effort which is needed 
to eventually support volume multiattach. John Garbutt and myself are 
working that from the Nova side, and John Griffith and Ildiko Vancsa 
from the Cinder side.

2. Continuing to unwind the CI dependencies on nova-network. I'm working 
on those efforts.

3. Get agreement on the discoverable capabilities API WG spec so we can 
work on implementing that in Pike.

4. Get the gate-tempest-dsvm-security job created which tests with 
Barbican as the key manager and it would verify signed images. This is a 
foundation on which we need to later build in new security-related 
features to Nova.


Finally, because of the short schedule, we agreed that there is just a 
single feature freeze which is the same as the rest of OpenStack for the 
Ocata release, which is January 26th. So there is no non-priority 
feature freeze for Ocata.



Matt Riedemann

More information about the OpenStack-dev mailing list