oslo.db model tests broken by alembic 0.8.4

[Ihar Hrachyshka]

neutron functional tests are broken [1] by new alembic 0.8.1 release
because it now catches some differences between models and migration
scripts that were not caught before (specifically, in foreign key
attributes).

[1]: http://logs.openstack.org/66/211166/2/gate/gate-neutron-dsvm-functional/867114c/

[Ihar Hrachyshka (ihar-hrachyshka)]

The global requirements fix for kilo: https://review.openstack.org/#/c/216328/

[Matt Riedemann (mriedem)]

http://logstash.openstack.org/#eyJzZWFyY2giOiJtZXNzYWdlOlwiQXNzZXJ0aW9uRXJyb3I6IE1vZGVscyBhbmQgbWlncmF0aW9uIHNjcmlwdHMgYXJlbid0IGluIHN5bmNcIiBBTkQgYnVpbGRfbmFtZTpcImdhdGUtbmV1dHJvbi1kc3ZtLWZ1bmN0aW9uYWxcIiIsImZpZWxkcyI6W10sIm9mZnNldCI6MCwidGltZWZyYW1lIjoiNjA0ODAwIiwiZ3JhcGhtb2RlIjoiY291bnQiLCJ0aW1lIjp7InVzZXJfaW50ZXJ2YWwiOjB9LCJzdGFtcCI6MTQ0MDQzNDEzMTkxMSwibW9kZSI6IiIsImFuYWx5emVfZmllbGQiOiIifQ==

[Henry Gessau (gessau)]

I don't understand why the gate pipeline is failing on that after it passed the check pipeline.

Looks to me like the gate does not always fast forward to the latest neutron code?

With the latest neutron code this error does not occur.

[Alan Pevec (apevec)]

@Ihar - is this https://bitbucket.org/zzzeek/alembic/issues/321 which was fixed in alembic 0.8.2 ?

[Alan Pevec (apevec)]

Archiving answer from IRC:

<ihrachys> apevec: https://bitbucket.org/zzzeek/alembic/issues/317/implement-autogen-detection-for-onupdate
<apevec> ok, so still an issue
<apevec> hm, says resolved?
<ihrachys> apevec: it's solved in 0.8.2. the problem is not the release itself, it's that neutron functional tests fail with it
<ihrachys> apevec: because now it catches some differences between models and alembic scripts that were not caught before
<ihrachys> and backporting the patch from master that syncs models is not an option because it introduces a new migration script
<ihrachys> there is no runtime issue per se, mere testing
<ihrachys> so if you are brave to run with smth that is not validated in gate, fine, you can do it with alembic 0.8.2+ for kilo too

[Sean Dague (sdague)]

This has popped back in Alembic 0.8.4 because of oslo.db tests

[Henry Gessau (gessau)]

Backporting https://review.openstack.org/258471 will address this.

We also need to fix the constraints for the functional job, since we should not have picked up 0.8.4 automatically.

[Henry Gessau (gessau)]

Comment #7 is invalid. Neutron stable jobs are not impacted by alembic 0.8.4 since they do not use variants.