[release][oslo] FFE request for Oslo Policy

Ghanshyam Mann gmann at ghanshyammann.com
Thu Apr 9 21:03:13 UTC 2020


 ---- On Thu, 09 Apr 2020 15:17:17 -0500 Sean McGinnis <sean.mcginnis at gmx.com> wrote ----
 >    > Hi,
 > >>   >
 > >>   > This is regarding the FFE request for policy for
 > >>   > - https://review.opendev.org/#/c/717879/
 > >>   >
 > >>   > This is part of introducing scope_type and new defaults in
 > >> policies. During
 > >>   > Nova implemtnation[1], we came across two things to solve before
 > >> we release
 > >>   > the new policies changes:
 > >>   > 1.  A lot of warnings as policy default are changing which end up
 > >> filling the logs (226 MB n-api logs today).
 > >>   > 2. Give operators an option to switch to the new system (scope
 > >> and new defaults) at the same time without overwriting the policy file.
 > >>
 > >> Basically two patches, I proposed those separately
 > >> 1.   https://review.opendev.org/#/c/717879/
 > >> 2.   https://review.opendev.org/#/c/717943/
 > >
 > > The first of these has already merged. Before I pull the trigger on
 > > the second, I want to make sure we're good on the FFE front. Given
 > > that these changes can't be backported and are fairly important for
 > > usability of the new policy work, I'm in favor of an FFE. Can someone
 > > from the release team weigh in too? Thanks.
 > >
 > This should be fine from the releases perspective. The more interesting
 > aspect is requirements.
 > 
 > Here are the repos that have oslo.policy as a dependency:
 > 
 > http://paste.openstack.org/show/791900/
 > 
 > So this impacts a lot more than nova.
 > 
 > That said, if the oslo team (I say with my shiny new oslo hat off) is OK
 > with the level of risk this introduces, then it should be fine to do
 > another release. We are two weeks past oslo lib freeze, and one week
 > past non-client lib freeze, so this is getting really late. But if we
 > are going to do it, it should be today so it doesn't end up any later
 > than it already is.

Impact on existing behaviour should be none as newly added flags are false by default and
until any users explicitly enable those flag they will not get any behaviour changes.

-gmann

 > 
 > Sean
 > 
 > 
 > 



More information about the openstack-discuss mailing list