Open Stack

On Fri, 12 Apr 2013 18:17:38 -0400
Sean Dague <sean at dague.net> wrote:

> On 04/12/2013 02:19 PM, Adam Young wrote:
> > The Keystone  SQL upgrade tests are not currently run against
> > PostgreSQL or MySQL.  As such, they have bit-rotted and no long run
> > successfully. Before we approave any more SQL upgrade script
> > changes we need to fix this.
> 
> It's pretty simple to address that, I think Chris Yeoh had some
> patches to bring in the nova db migrations testing framework so that
> those would be tested on every commit.
> 
> Maybe we should just make it a priority to get those in?

The initial patch for the framework is here:

https://review.openstack.org/#/c/23660/

When enabled in the gate it will automatically run against PostgreSQL,
MySQL and SQLite so it will stop the bitrot in the future. It was last
rebased just over a couple of weeks ago
and auto-abandoned, but I've just restored it if anyone wants to review
it.

Note that it depends on:

https://review.openstack.org/#/c/25065/

which explicitly closes session in db migration 022. Otherwise the new
test db framework tests fail as there are left over connections to the
db when the test cleanup attempts to remove the temporary database
(only an issue with PostgreSQL as the backend).

The initial framework changeset only includes a testcase for the 002
migration, but there is a series of changesets which implement testcases
and remove redundant ones from test_sql_upgrade in the queue for db
migrations up to 019. Though they probably need rebasing again and need
to restored as they timed out with the others.

Adam - please let me know if you believe there is anything else I need
to change with the framework or tests.

Regards,

Chris
-- 
yeohc at au1.ibm.com