Overall plan and status tracking document: https://etherpad.openstack.org/p/stx-openstack-patch-refactoring

· Details: https://docs.google.com/spreadsheets/d/1udAtEpQljV2JZVs-525UhWyx-5ePOaSSkKD1CS27ohU/edit?usp=sharing

Story boards: https://storyboard.openstack.org/#!/story/list?status=active&tags=stx.distro.openstack&project_group_id=86

10/8/2019 meeting

· Upgrade to Train - Shuicheng suggests deferring Keystone upgrade until the CentOS RPM is available.  Is this OK, will we hit compatibility issues?  

• Still planning to wait for CentOS - could be a month or two - will run tests to verify functionality.  Risk should be low and we can always build the sRPM ourselves.  Dean says that the API changes are mainly in changes we pushed.

· Test plan update  - Yong

• Testing for this? https://governance.openstack.org/tc/goals/selected/train/ipv6-support-and-testing.html.  Dean has been testing ipv6 OpenStack but is seeing problems outside of O.S. that make running only ipv6 hard - e.g. github & gitlab don't support it at all.  Some customers want ipv6 only.   There is a setup in the SH lab for ipv6 we could use to start this.  Yong thinks we should start testing this feature sooner rather than later.  JC (with help from Elio who has tested this in the past) could use the SH setup to run a few basic tests eg. Sanity.

· Train migration update https://storyboard.openstack.org/#!/story/2006544

• Yong: Regarding Train upgrade, we are making some progress (docker image creation is done, a helm issue has been addressed and the manifest is ready) and very likely we will work out an EB for testing by Oct 11th, next Friday. And will try to catch up making patches ready before Feature Freeze Milestone Oct 16.  The patches are small and easy to review.  

· Opens

• Nova patch to detect failed VMs - big performance increase - do we want to push upstream?  Carry the patch?

§ Patch is visible here: https://gist.github.com/VictorRodriguez/e137a8cd87cf821f8076e9acc02ce195

§ Dean: Nova might not be the best place to land this feature.  Nova isn't designed (now) to support this and is unlikely to accept the changes - indeed they have already been rejected.  If we need sub-second response times for detecting failed VMs we should find a place to land this somewhere else.  Could also be added to libvirt (maybe – might need libvirt changes).  Will require some architecture work to figure out the best path forward.

• Will shorten this meeting to 30 minutes going forward so Frank can have the bottom of the hour time slot for the container meeting.