[openstack-dev] What's happening with stable release management?

Thomas Goirand zigo at debian.org
Fri Aug 8 17:28:50 UTC 2014


For updating keystone from 2014.1.1 to 2014.1.2, I had to:

- Upgrade oslo-config from 1.2.1 to
- Upgrade oslo.messaging from 1.3.0~a9 to
- Upgrade python-six from 1.6 to 1.7
- Upgrade python-pycadf from 0.4 to 0.5.1
- Add python-ldappool
- Add python-oslo.db
- Add python-oslo.i18n
- Add python-keystonemiddleware, which needs python-crypto >= 2.6
(previously, 2.5 was enough)

So, we have 5 major Python module upgrades, and 4 completely new
libraries which were not there in 2014.1.1. Some of the changes aren't
small at all.

I'm sure that there's very valid reasons for each of the upgrades or
library addition, but I don't think that it is overall reasonable. If
this was to happen during the freeze of Debian, or worse, after a
release, upgrading all of this would be a nightmare, and I'm sure that
the Debian release team would simply refuse.

Should I assign myself to program a robot which will vote -1 on all
change on the stable/Icehouse global-requirements.txt file? Or is sanity
still possible in OpenStack? :)

It is my opinion that we need to review our release process for the
stable releases, policy for requirement changes, and need to adopt a way
more conservative attitude.


Thomas Goirand (zigo)

More information about the OpenStack-dev mailing list