candice

This is an early release. Some features are not yet fully implemented.

Candice

Cloud-Native Network Device Configurator.

Get It

Rationale

Managing network services requires configuring their network functions. Whether those reside in clouds (CNFs and VNFs) or in boxes (PNFs) they may require sending the configuration data via NETCONF or similar protocols.

This can be subtle work, as much of this data is contextual and conditional, e.g. you must query for what network interfaces are available and what features they have before setting them up to achieve the required functionality. Often this back-and-forth is handled by a controller or orchestrator where you can design "workflows", "playbooks", "recipes", etc., in the context of the network service more generally.

The problem is that as our deployments grow into thousands and tens of thousands of sites, with hundreds of thousands of network functions, these big and faraway orchestrators struggle to handle the scale.

Candice can make this easier by delegating the function's configuration work to its site, indeed it should run within CNF's cluster. Work is encapsulated into goal-oriented "tasks", which can be understood as "micro-workflows" within the general orchestration context. Each task can make many different NETCONF calls, even on several devices, and can rely on relevant state managed locally instead of at the faraway orchestrator. The orchestrator can then just tell Candice to run the whole task with custom inputs and outputs.

How It Works

Candice is a Kubernetes operator that:

1. lets you to create "Device" custom resources, which contain connectivity information (addresses, protocols, credentials) for a device. Currently the NETCONF protocol is supported.
2. lets you attach "tasks" to the operator. A task is a Python program that handles a specific configuration workflow on one or more devices. Tasks can accept arbitrary inputs and emit arbitrary outputs, allowing you to customize them as necessary for integration with management and orchestration solutions, such as MANO.
3. provides a rich Python API to make it easier to write tasks, e.g. setting up NETCONF connectivity and allowing you to read and write configuration data. Also included is persistent transactional state, a "scratch" area that you can use for whatever state is necessary for your tasks, including for tasks to share state with each other.

Here are examples of a few tasks.

Usage example:

candice device create router --host router.mynamespace.svc.cluster.local:830

candice task set router about --file=examples/router/tasks/about.py
candice task set router set-hostname --file=examples/router/tasks/set_hostname.py

candice task run router set-hostname --input=hostname=myhost
candice task run router about