goSDN
is a prototypical approach to build a model driven multi-vendor SDN controller.
This repository contains submodules, therefore it must be cloned with
git clone --recurse-submodules git@code.fbi.h-da.de:danet/gosdn.git
or if you have cloned the repository without --recurse-submodules
simply run:
git submodule update --init --recursive
to get all submodules.
You can find a detailed manual to install all necessary tools at our Wiki.
There you can also find some tutorials to get to know the SDN controller:
- Lab00 education, which is used in education
- Lab00, which uses servers with gNMI targets instead of Arista switch images (needs less storage than Lab01)
- Lab01, which uses servers with Arista switch images (needs less storage than Lab01)
Table of Contents
Overview
goSDN
is also an application that will allow you to manage your multi-vendor
network using one unified controller.
goSDN
provides:
- Model driven device/network element representation
- Native multi vendor support
- Multi controller environments
Example
A simple showcase how the controller can be adressed after
make containerlab-start
is shown below:
Structure
- The
api
is a representation of thecontrollers
Northbound-Interface and can be used to communicate with thecontroller
. - The
cli
is the CLI to manage thecontroller
. -
csbi
is the implementation of Containerised-Southbound-Interfaces (based on the idea and the proof of concept of Manuel Kieweg). Allowing to request capabilities of MNEs and generate a containerised Southbound-Interface based on them. They are currently unsupported. -
controller
represents thegoSDN-controller
.
Concepts
The goSDN
controllers core - also called nucleus
- is a lightweight library
that manages principal network domains and provides southbound interface
operations for managed network elements.
In addition, we provide a simple Northbound-API for the controller right here.
Principal Networking Domain (PND)
The PND is the single source of truth within a network. Its state is held and maintained by the controller. Any configuration of an MNE has to be applied by the PND.
Managed Network Element (MNE)
Any network element directly configured by goSDN
Launch goSDN Controller local
In this chapter, you learn how to launch the goSDN controller.
Firstly, make sure that you're located in the root directory of gosdn.
goSDN
provides a Makefile
for all common use cases.
# build the application files for goSDN, cSBI orchestrator, gosdnc (CLI).
make build
# build the Dockerfiles for goSDN, cSBI orchestrator, gosdnc (CLI).
make containerize-all
Depending on how you want to use the controller, you might need to edit the example config files and provide the path to this or a new config file. This can be done by providing setting the config flag when starting the controller.
--config
In case you want to use specific services with the controller without setting them up manually you can use the provided docker compose file. More information about that can be found under Additional services.
Now you can start goSDN
locally by moving to the artifacts
folder and
running ./gosdn
from the shell:
./gosdn
Getting Started
If you want to use the the playground you have to make sure you have containerlab installed on your system.
Playground
With the help of containerlab we provide simple test environments to play around with.
The environment Lab 01 contains two Arista cEOS, a goSDN, a gNMI target, a plugin-registry, a MongoDB and a RabbitMQ.
The environment Lab 00 contains two gNMI targets used as switches, s goSDN, two Linux servers, a plugin-registry, a MongoDB and a RabbitMQ. Lab 00 is lighter than Lab 01.
If you're a member of the danet group you should have access to the containers repo. Don't forget to
sudo docker login registry.code.fbi.h-da.de
(sudo
is necessary sincecontainerlab
has to be run with it and therefore a normaldocker login
wouldn't be recognized by it.)If you're no member of the danet group you have to create an account at Arista and download the Arista cEOS image by yourself. Don't forget to change the image name (line
11
in thegosdn.clab.yml
file) to the local one you've downloaded.
# starts the containerlab topology which contains two Arista cEOS, an cSBI orchestrator, a goSDN controller and gNMI target.
make containerlab-start
# stop the containerlab topology
make containerlab-stop
Additional services
There are optional and mandatory services that can be used with the controller. They can be started via the provided docker-compose.yml
:
docker compose up -d
Otherwise all the necessary services are included in the provided environments like containerlab. Within other use cases, they can also be configured by providing specific parameters in the controller config files. See Configuration file for information about the parameters. The services are:
- RabbitMQ message-broker: uses AMQP to provide a pub-sub mechanism. This is part of the event system used in goSDN. This server is optional.
- Databases: The controller can be used with a file store system or by providing an external database. Currently, there is only support for MongoDB. It is not mandatory to use a database.
- Plugin-registry: The plugin-registry provides plug-ins to the controller, which are used to handle different YANG files used within the communication with MNE via gNMI.
Configuration file
goSDN can be configured by using providing a config file. Examples can be found in controller/configs
. These examples contain different pre-defined configurations for different environments. There's no need to touch the config files when using provided environments like starting the containerlab via makefile.
The following parameters can be set:
For goSDN:
-
security
: the security level of the controller.-
'secure'
mode does use the built-in authentication and RBAC mechanisms (recommended) -
'insecure'
does not use these (only use for testing purposes). Uses default if not set.
-
-
socket
: the desired port, where to reach the controller -
basepnduuid
: an initial ID of a PND within the network, needs to be a UUID -
config
: path to this configuration file, so change can be written back to the file -
gnmisubscriptionspath
: path to the file providing the YANG paths the controller should subscribe to in each network element existing in its storage -
defaultjwtduration
: duration of how long session tokens will be valid, provide an integer for the desired hours, example: 24 for tokens that should be valid for one day -
log-level
: set to'debug'
for a better level of information -
help
: print the description of all available flags
For the storage system:
-
databaseconnection
: address of the database, only add this parameter if you want to use a database. Currently only mongoDB is supported. -
filesystempathtostores
: set this path if the controller should use only the file store system, does not support all entities currently.
Note: if the database connection is set as well, this parameter will be ignored. -
plugin-folder
: path to where the plugins are/should be stored.
For the event system (RabbitMQ broker):
-
amqpprefix
: the URI prefix that is needed to connect to the RabbitMQ broker. -
amqphost
: ip adress of the broker -
amqpport
: port of the broker -
amqpuser
: user name of credentials to access the broker, default value isguest
-
amqppassword
: user password of credentials to access the broker, default isguest
For the plugin registry:
-
plugin-registry
: ip adress and port of the plugin-registry
CLI
Information about how to use the CLI is provided in the cli
folder, see here.
CI Status
Master |
---|
Development Tutorial
To help with developing there is a small script available, which helps to create the complete dev environment.
For it to work, you need go
, docker
, docker compose
and containerlab
.
For a quick start simply use make start-dev-env
to start everything. This can take a while. Afterwards, you see the output of goSDN, to stop it hit Ctrl+C. To stop everything else use make stop-dev-env
.
If you want more control over your environment, you can use the manage_virt_env.sh
script directly.
It works like this, if you are in the projects root folder:
./scripts/manage_virt_env.sh --mode $(start || stop) --topology $PATH_TO_TOPOLOGY --SDNCONFIG $PATH_TO_SDNCONFIG --keepdb (optional)
mode
defines if it should start or stop the environment, while the two commands --topology
and --sdnconfig
are used to provide the desired files.
The optional flag keepdb
is used if you want to create the environment without reloading the SDN config file, which is handy if you just want to restart the environment. For every command there is a shortcut available (e.g. -t for --topology, etc.).
Example usages:
./scripts/manage_virt_env.sh --mode start --topology dev_env_data/clab/basic_two_aristas.yaml --sdnconfig dev_env_data/sdn/basic_two_aristas.json
./scripts/manage_virt_env.sh --mode start --topology dev_env_data/clab/basic_two_aristas.yaml --keepdb
./scripts/manage_virt_env.sh --mode stop --topology dev_env_data/clab/basic_two_aristas.yaml
Backup and restore your mongodb
If you are using the mongodb provided via the docker-compose.yaml
, you can easily back up and restore your data via two scripts.
./scripts/backup_mongo_volume.sh $(suffix_of_backup)
./scripts/restore_mongo_volume.sh $(suffix_of_backup)
Keep in mind that a restore will restart the mongodb container.
Developing SDN applications
To develop applications, we provide a framework that can be found here within this repository.
The framework provides some basic code to easily setup the subscription to the event system of the controller, so that your application can get notified via the events published by the RabbitMQ server about information you would like to receive. This includes functions to register and subscribe to the used server. The information you receive includes changes on network elements like if a value of a YANG path changes or other changes like additions or deletions of entities.
Examples where the application framework is used can be found here.