[openstack-dev] [legal-discuss] [Marconi] Why is marconi a queue implementation vs a provisioning API?

Kurt Griffiths kurt.griffiths at rackspace.com
Thu Mar 20 16:29:22 UTC 2014

> The incorporation of AGPLv3 code Into OpenStack Project  is a
>significant decision

To be clear, Marconi does not incorporate any AGPL code itself; pymongo is
Apache2 licensed.

Concerns over AGPL were raised when Marconi was incubated, and I totally
respect that some folks are not comfortable with deploying something like
MongoDB that is AGPL-licensed. Those discussions precipitated the work we
have been doing on SQLAlchemy and Redis drivers. In fact, the sqla driver
was one of the graduation requirements put in place but the TC. Now, if
people want to use something else than what the Marconi team is already
working on, we are more than happy to have them contribute and help us
evolve the driver interface as needed.

On the subject of minimizing the number of different backends that
operators need to manage, some relief may be found by making the backends
for projects more customizable. For example, as we move more projects to
using the oslo caching library, that will give operators an opportunity to
migrate from memcached to, say, Redis. And if OpenStack Service X
(Marconi, for example) supports Redis as a backing store, now the operator
can reuse their Redis infrastructure and know-how.

The software industry has been moving towards hybrid NoSQL+SQL
architectures for a long time now, in order to create best-fit solutions;
I think we’ll see more OpenStack projects following this model in the
future, not less, and so we need to work out a happy path for supporting
these kinds of operating environments.


More information about the OpenStack-dev mailing list