Juju 2.x

Proposer

Mark Shuttleworth Marco Ceppi

Type

Feature

Target MDG/TF

(Optional)

VNF Configuration, DevOps

Description

OSM R0 uses Juju 1.x for modelling VNFs. Also, OSM itself will be charmed, making it much easier to deploy and operate both for evaluation and production.

In Juju 2.x we gain a number of valuable capabilities:

  • Multi-model controller. A single instance of Juju can manage multiple models. This means that RIFT.io will be able to stand up multiple independent scenarios using a single Juju instance, rather than needing separate Juju controllers per scenario. This will reduce the footprint of OSM and improve the operations story for production users.
  • Workload status. VNF charms can provide a clear status which RIFT will be able to interpret and present to the user. VNF vendors will find it easier to express the status of their VNFs in a complex topology.
  • LXD containers. The "local" provider, which is used for devops iteration, moves to LXD which is a full hypervisor and offers much faster and more flexible local deployment operations. VNF endors will have a far more efficient test and dev experience providing OSM VNFDs.
  • Multi-user controller. A single Juju controller can track multiple users, who have independent access to particular models managed by that controller. For OSM to enter production usage we need to be able to offer audit and ops access at each level of the stack, and Juju 2.x enables this in a PCI-DSS compliant fashion.
  • Resources. A charm can now specify a set of resources that will be delivered to the charm at runtime. Common examples would be the charm's preferred JVM, or a docker image. Juju 2.x mediates access to these resources for efficiency, the net result is a greatly improved VNFD experience for VNF vendors.

Juju 2.0 is currently in late beta, APIs are finalised, with a final release expected in July, so integration work can commence without having to wait.

The work involved in this change should come about naturally - initial Juju 1.x integration with RIFT was straightforward and Juju 2.x presents a cleaner API, so we do not anticipate challenges on this front. The Canonical team is happy to help the RIFT team with this update as part of our joint work on OSM.

Demo or definition of done

This change would be somewhat invisible to end-users. It will substantially improve the experience of VNF vendors because Juju 2.x charms have access to important new capabilities.