Latest revision as of 09:51, 15 January 2018

Containers in Magnum clusters are created only using the native container orchestrator API. Please refer to:

• https://docs.openstack.org/magnum/latest/user/index.html#native-clients
• https://docs.openstack.org/magnum/latest/install/launch-instance.html

NOTICE: This is a work-in-progress draft. Our community is working on refining this draft here: https://etherpad.openstack.org/p/magnum-native-api

Magnum's Native Container APIs

Magnum bays support native API access as well as a more limited form of wrapped access to that same functionality through a /container resource in the Magnum API. This means that, depending on which bay type you select: Docker Swarm, Kubernetes, or Apache Mesos, you get the API that comes with that container orchestration engine. For example, if I create a Bay in Magnum using the Docker Swarm bay type, I will have an endpoint for the Docker API that I can interact with using the native docker CLI tools.

Project Position on Native APIs

Magnum's contributors recognize the native API approach is preferred, as it offers users a more complete set of features and capabilities, and those can be updated without waiting for Magnum to re-implement them. It also allows for use of native tooling, which is the preferred user experience.

Use Cases for Magnum's "container" Resource

We are seeking production use cases for the /container resource offered in the Magnum API. If you have one, please identify it here:

• Deploy containers from other OpenStack services. Expecting services to abstract every possible supported bay type is not scalable. A more precise example, I'd like trove to deploy DBs on containers and to do that, I'd like it to use Magnum's containers API. (flaper87)