api db_sync fails on upgrade due to duplicate migration files

[Steven Hardy]

I'm testing upgrades on TripleO, and hit this problem, discussion with owalsh indicates it may be a nova bug related to backport migration numbering:

14:44 < owalsh> shardy: it's a nova bug, 028 needs to be renamed to 021_build_requests_instance_mediumtext.py in ocata &
                master

This is the error:

"ScriptError: You can only have one Python script per version, but you have: /usr/lib/python2.7/site-packages/nova/db/sqlalchemy/api_migrations/migrate_repo/versions/021_build_requests_instance_mediumtext.py and /usr/lib/python2.7/site-packages/nova/db/sqlalchemy/api_migrations/migrate_repo/versions/021_placeholder.py"], "warnings": []}

I'm testing this version:

[root@overcloud-controller-0 ~]# rpm -qf /usr/lib/python2.7/site-packages/nova/db/sqlalchemy/api_migrations/migrate_repo/versions/021_placeholder.py
python-nova-15.0.0-0.20170215034806.bdeb05d.el7.centos.noarch

So upgrading to:
https://github.com/openstack/nova/commit/bdeb05dfb0f727654ac0b0bae14341fd87b5cbb7

From stable/newton commit:
https://github.com/openstack/nova/commit/c6743ca709d45334cf25332aa834f86a9d91f1a5

[root@overcloud-controller-0 ~]# rpm -qa --last | grep python-nova
python-nova-15.0.0-0.20170215034806.bdeb05d.el7.centos.noarch Wed 15 Feb 2017 07:02:12 PM UTC
python-nova-14.0.4-0.20170117154931.c6743ca.el7.centos.noarch Mon 23 Jan 2017 01:45:50 PM UTC

[Matt Riedemann (mriedem)]

This doesn't really make any sense, it's 28 in ocata:

https://github.com/openstack/nova/tree/stable/ocata/nova/db/sqlalchemy/api_migrations/migrate_repo/versions

And 21 in newton:

https://github.com/openstack/nova/tree/stable/newton/nova/db/sqlalchemy/api_migrations/migrate_repo/versions

Plus grenade would be failing if we screwed this up...

[Matt Riedemann (mriedem)]

Dan pointed out to me in IRC the problem is 021_placement in stable/ocata conflicts with 021 from stale/newton.

[Steven Hardy (shardy)]

Ok so further testing this may be packaging related, as I ended up with two versions of python-nova installed:

[root@overcloud-controller-0 ~]# rpm -ql python-nova | grep migrate_repo | grep "py$" | grep 021
/usr/lib/python2.7/site-packages/nova/db/sqlalchemy/api_migrations/migrate_repo/versions/021_build_requests_instance_mediumtext.py
/usr/lib/python2.7/site-packages/nova/db/sqlalchemy/api_migrations/migrate_repo/versions/021_placeholder.py
[root@overcloud-controller-0 ~]# rpm -qa | grep python-nova
python-nova-15.0.0-0.20170215034806.bdeb05d.el7.centos.noarch
python-nova-14.0.4-0.20170117154931.c6743ca.el7.centos.noarch
[root@overcloud-controller-0 ~]# rpm -qf /usr/lib/python2.7/site-packages/nova/db/sqlalchemy/api_migrations/migrate_repo/versions/021_build_requests_instance_mediumtext.py
python-nova-14.0.4-0.20170117154931.c6743ca.el7.centos.noarch
[root@overcloud-controller-0 ~]# rpm -qf /usr/lib/python2.7/site-packages/nova/db/sqlalchemy/api_migrations/migrate_repo/versions/021_placeholder.py
python-nova-15.0.0-0.20170215034806.bdeb05d.el7.centos.noarch

It's not yet clear how that happened, but I'll remove nova from this since it appears to be an issue specific to package updates and/or the tripleo upgrade workflow (in particular I think we can end up in this state if an upgrade fails during the package upgrade, then the upgrade is re-attempted).