Open Stack

On Fri, Jan 24, 2014, at 12:55 PM, Clint Byrum wrote:
> Excerpts from Tim Bell's message of 2014-01-24 11:55:02 -0800:
> > 
> > This is exactly my worry... at what point can I consider moving to MariaDB with the expectation that the testing confidence is equivalent to that which is currently available from MySQL ?
> > 
> > The on-disk format is not so much a concern but there are many potential subtle differences in the API which can occur over time such as reserved words or recovery handling from certain errors.
> > 
> > Currently, my 'operational' hat says stay with MySQL. My 'community' hat direction is less clear. Given that Ubuntu and Red Hat are not agreeing on the direction makes this likely to be an extended uncertainty.
> > 
> > Maybe two yes/no answers would help....
> > 
> > 1. Is there any current blocking gate tests with MariaDB ?
> > 2. Is there a plan to change this at a future release ?
> > 
> > If the answer to 1 is no, the operational hat says to stay on MySQL currently.
> > 
> > If the answer to 2 is yes, we should be planning to migrate.
> 
> I think another question you may want to ask is "are there gate blocking
> tests on RHEL/CentOS/etc.?"
> 
> _______________________________________________
> OpenStack-dev mailing list
> OpenStack-dev at lists.openstack.org
> http://lists.openstack.org/cgi-bin/mailman/listinfo/openstack-dev

There is actually, but only for python2.6 unittests run on CentOS 6.5.
All of the tempest tests run on Ubuntu Precise with most jobs using
MySQL and a few using PostgreSQL. None of the tests use MariaDB today.
There were folks working on getting Fedora nodes into nodepool to do
testing of some sort on Fedora (which would be MariaDB aiui), but ran
into trouble around nodepool's privilege expectations (something about
sudo needing a proper tty and paramiko not being able to nicely provide
one). We should pick that work back up again, but Infra has been swamped
recently.

If we do end up adding MariaDB integration (tempest) tests I would not
add another set of tests. Instead we should probably swap out MySQL and
use MariaDB in one of the existing tests.

Clark