Hi all, until this point we have been able to maintain 1:1 mapping from Fedora to OpenStack releases [1] which has some advantages for package maintainers in RDO, rdopkg info output is nicely symmetrical and predictable: $ rdopkg info RDO releases & repos: kilo (Fedora 23/rawhide) fedora-22 built in koji/master from master branch fedora-21: symlink to fedora-22 repo (don't submit fedora-21 updates) epel-7 built in cbs/cloud7-el7 from master branch juno (Fedora 22) fedora-21 built in koji/f22 from f22 branch fedora-20: symlink to fedora-21 repo (don't submit fedora-20 updates) epel-7 built in copr/jruzicka/rdo-juno-epel-7 from f22 branch icehouse (Fedora 21) fedora-20 built in koji/f21 from f21 branch epel-6 built in copr/jruzicka/rdo-icehouse-epel-6 from f21 branch epel-7 built in copr/jruzicka/rdo-icehosue-epel-7 from f21 branch But OpenStack releases are coming up quickly, upstream stable/juno will be already in support phase II by the time Fedora 22 is released[2] We have a window of opportunity to push Kilo to Fedora 22 before it freezes for Beta end of the month[3] Advantage would be that we'll get fresh OpenStack release in Fedora 22 at the cost of skipping one OpenStack release in Fedora and dropping RDO Juno for Fedora 21. I'd like to hear from users if that's acceptable. Assumption here is that Fedora users are mainly developers who are already following Kilo development while stable production is on EL platforms. We would keep RDO Juno for EL7 (and EL6 which is in the works) by moving current f22 dist-git branches to github/openstack-packages rpm-juno branches and eventually to CentOS CloudSIG dist-git when that's ready. Kilo3 import would be then finished on Fedora master and merged to f22 early next week. Please provide feedback quickly if this plan is acceptable, we need to execute it before March 31st.

[1] https://fedoraproject.org/wiki/OpenStack#OpenStack [2] https://wiki.openstack.org/wiki/StableBranch#Support_phases [3] https://fedoraproject.org/wiki/Releases/22/Schedule

Did we had an outcome of this?

Is this still necessary or can I propose to drop Fedora from the official Installation Guide? This would safe us a lot of time because we do not have to test for Fedora any longer and I do not have to update the Installation Guide for Fedora 21.

-----BEGIN PGP SIGNED MESSAGE----- Hash: SHA1 On 04/13/2015 12:46 PM, Alan Pevec wrote: > 2015-04-10 19:23 GMT+02:00 Matthias Runge <mrunge(a)redhat.com&gt;: >> Did we had an outcome of this? > > Status is that we missed F22 Beta freeze on Mar 31, now fesco > freeze exception is required and Kilo import is still not completed > in Rawhide. I'm also having second thoughts after thinking about > Ihar's feedback last week on IRC where he questioned whether the > price of maintaining openstack-* packages in stable Fedora is worth > it when nobody is actually using them. Until proven otherwise, > we're assuming typical OpenStack user on Fedora is interested in > development and wants only latest packages, so CVE fixes and > upstream stable point release updates are wasted. Production users > are expected to run OpenStack on EL platform, which we provide > through RDO. > > Instead of rushing Kilo into F22, I was thinking about the > following model: (1) keep latest OpenStack release in Fedora > Rawhide, starting from milestone2. Delorean stays RDO Trunk as a > place for tracking changes on OpenStack master branch. (2) > deprecate openstack-* in released Fedora to avoid maintenance > overhead for no use (3) provide Fedora Rawhide builds for current > and previous Fedora in RDO repo e.g. Kilo would be available for > F22 and F21 through RDO (4) document that for production EL > platform is recommended > > We cannot retire packages in released Fedora, so I'm not sure > about the process for (2). f20 branches should definitely be closed > since Havana is EOL upstream and in RDO, so there will be no > further updates. If there are security bugs, we should have an ability to patch those. So -1 to retiring branches. I'm fine supporting those packages that are already released in current releases (f20 and f21) while those releases are alive, if that's a temporary transition thing. > When I tried fedpkg retire in openstack-keystone f20 branch, > "dead.package" commit in dist-git worked[*] but then pkgdb step > failed. Ideally, we would push an "empty" update which would warn > installed user base that they're running EOLed packages and should > migrate to RDO but I'm not sure there's such mechanism in Fedora > Bodhi. I'd appreciate any and all feedback on the above plan, > please point out the holes it might have. If the plan makes sense, > F22 would stay Juno and it would be irrelevant which OpenStack > release ends up in Fedora >= 23. > > > Cheers, Alan > > [*] > http://pkgs.fedoraproject.org/cgit/openstack-keystone.git/commit/?h=f20&a... > > _______________________________________________ Rdo-list mailing > list Rdo-list(a)redhat.com > https://www.redhat.com/mailman/listinfo/rdo-list > > To unsubscribe: rdo-list-unsubscribe(a)redhat.com > _______________________________________________ Rdo-list mailing list Rdo-list(a)redhat.com https://www.redhat.com/mailman/listinfo/rdo-list To unsubscribe: rdo-list-unsubscribe(a)redhat.com

I'd even go the route to push horizon in kilo version to f22, even if the rest of the stack will stay on Juno. In theory that should work, as it did on my dev platform (using git checkout and a remote RHOS-6 installation).