Open Stack

On Fri, 2014-03-28 at 18:16 +0000, Duncan Thomas wrote:
> On 28 March 2014 14:38, Joshua Harlow <harlowja at yahoo-inc.com> wrote:
> > An idea that might be good to do is to start off using SQLite as the
> > taskflow persistence backend.
> >
> > Get that working using SQLite files (which should be fine as a persistence
> > method for most usages) and then after this works there can be discussion
> > around moving those tables to something like MySQL and the
> > benefits/drawbacks of doing that.
> >
> > What do u think?
> >
> > Start small then grow seems to be a good approach.
> 
> I'd say at tables to the existing database first, it adds the smallest
> number of new moving parts, then look at a new database if and only if
> there seems to be evidence that one is needed. As a systems operator,
> it is hard enough managing one db let alone N local dbs...

My suggestion would be to add a new CONF option
(persistent_db_connection?) that would default to the value of
CONF.db.connection). That way you default to using the same DB as the
main Cinder tables, but automagically provide the deployer with the
ability to handle the taskflow tables separately if they want.

best,
-jay