nova-docker driver lets Nova, the OpenStack Compute service,
spawn Docker containers instead of hypervisor-based servers. For
certain workloads, this leads to better resource utilization than you
would get with a hypervisor-based solution, while at the same time
givin you better support for multi-tenancy and flexible networking
than you get with Docker by itself.
The Docker driver wiki was mostly sufficient for getting the
nova-docker driver installed in my existing OpenStack deployment,
although I did make a few small changes to the wiki to reflect
some missing steps. Other than that, the installation was relatively
simple and I was soon able to spin up Docker containers using
nova boot ...
The one problem I encountered is that it is not possible to pass
environment variable to Docker containers via the
driver. Many existing images (such as the official MySQL image)
expect configuration information to be passed in using environment
variables; for example, the
mysql image expects to be started like
docker run --name some-mysql \ -e MYSQL_ROOT_PASSWORD=mysecretpassword -d mysql
I have proposed a patch to the
nova-docker driver that permits
one to provide environment variables via the Nova metadata service.
With this patch in place, I would start the
mysql container like
nova boot --image mysql --flavor m1.small \ --meta ENV_MYSQL_ROOT_PASSWORD=mysecretpassword \ some-mysql
That is, the driver looks for metadata items that begin with
and transforms these into Docker environment variables after stripping
ENV_ from the name.
While this patch works great in my testing environment, it’s unlikely to get accepted. Generally, the metadata provided by Nova belongs to the tenant and is not meant to be operationally significant to the compute driver itself.
It sounds as if there is a lot of work going on right now regarding container support in OpenStack, so it is very likely that a better solution will show up in the near future.
In the absence of that support, I hope others find this patch helpful.