On Thu, Feb 20, 2020 at 8:49 PM Wesley Hayutin <whayutin@redhat.com> wrote:


On Thu, Feb 20, 2020 at 11:56 AM Mike Burns <mburns@redhat.com> wrote:
So, if I understand this right...

* We're keeping the centos7 and centos8 versions of Ussuri consistent
* There is no actual build of the latest commits built on either centos7 or centos8
* There is no possible way to have an Ussuri release on Centos7

Assuming the above is true, I would say to turn off all the centos7 stuff and focus all effort on getting Ussuri latest built on centos8.  There does not appear to be any value in continuing work on centos7 if we can't actually deliver it at the end of the day.

Mike

Agreeing and building on what Mike said.. this certainly raises the priority on CentOS-8.  There are still tripleo patches incoming that will not be pinned ( I think )
Having a better understanding here of what has been pinned and what is in progress w/o consulting rdo-info all the time would be handy.


Yes, tripleo patches are not pinned and should keep bulding in current repo..

Currently pinned packages because of py2 issues:

- nova
- keystone
- cinder
- glance
- neutron
- neutron-lib
- neutron-vpnaas
- networking-generic-switch
- horizon
- heat-dasboard
- murano-dashboard
- octavia
- mistral
- ironic
- ironic-staging-drivers
- ironic-python-agent
- hardware
- kolla
- senlin
- tobiko
- rally
- heat-tempest-plugin
- zaqar-tempest-plugin
- mistral-tempest-plugin

Also, update of oslo-libraries to latest releases is blocked so we are still using previous ones.
 
I think the check, gate and promotion jobs for centos-7 ussuri need to stay in place, however they should just always pass which would lower the amount of work w/ regards to centos-7 but not eliminate it.   It could also be the case that tripleo breaks itself from time to time, but I'm sure Emilien would never let that happen as he's a robot and not human.


There should be no problem with check jobs for c7, that will keep using latest promoted content + current tripleo and family. TripleO patches requiring changes in pinned projects will fail in CentOS7.

WRT promotion based on consistent link, as we get consistent link stuck by unpinning packages to get them updated in CentOS8, i'm not sure if periodic pipeline will keep running or it detects the link is still pointing to the same hash and will skip it.
 
I digress 
 

On Thu, Feb 20, 2020 at 9:05 AM Alfredo Moralejo Alonso <amoralej@redhat.com> wrote:

Hi,

I'd like to open a discussion about the status of RDO Ussuri repositories on CentOS7.

As you know RDO and upstream teams (kolla, puppet-openstack, TripleO, TripleO CI, etc...) have been working to switch to CentOS8 during last few weeks.

In order to make the transition easier from CentOS 7 to CentOS 8, RDO is still maintaining Trunk repos consistent for both CentOS 7/Python 2 and CentOS 8/Python 3. As OpenStack projects have been dropping support for Python 2, we've started pinning them to the last commit working with Python 2[1], we were expecting that transition will finish soon but it's still going on. Over time, the number of pinned packages has been growing including services and Oslo libraries where we can't follow upper-constraints anymore[2]. Recently, Kolla has removed support for CentOS 7 so i doubt it makes sense to keep pinning packages to keep RDO Trunk consistent artificially and continue running promotion pipelines on a repo with so many outdated packages. Also, pinning these projects makes that changes needed for CentOS 8 will not be in RDO and would need to be backported manually to each package. My proposal is:

- Unpin all packages in Ussuri to follow master trunk, or versions in upper-constraints (for clients and libraries).
- RDO Ussuri on CentOS 7 repo consistent link will not move anymore (so no more promotions based on it).
- We will keep running centos7-master DLRN builder, so that packages still builing with Python 2 will be available in current repo [3] to be used by teams needing them until migration to CentOS 8 is finished everywhere.
- Projects which already have CentOS 8 jobs gating in master branch can remove CentOS 7 ones.

We understand this can add some pressure on moving to CentOS8 to the teams working on it, but I'd say it's already a priority and it's justified at this stage.

What do you think about this plan?, is there any reason to keep CentOS 7 artificially consistent and promoting at this point of the transition to CentOS 8?

Best regards,

Alfredo

_______________________________________________
dev mailing list
dev@lists.rdoproject.org
http://lists.rdoproject.org/mailman/listinfo/dev

To unsubscribe: dev-unsubscribe@lists.rdoproject.org
_______________________________________________
dev mailing list
dev@lists.rdoproject.org
http://lists.rdoproject.org/mailman/listinfo/dev

To unsubscribe: dev-unsubscribe@lists.rdoproject.org