Skip to content

Latest commit

 

History

History
118 lines (87 loc) · 4.29 KB

README-openvidu.md

File metadata and controls

118 lines (87 loc) · 4.29 KB
Raw

openvidu-livekit

This fork of livekit enables the storage of livekit events and stats.

By default it stores them in a MongoDB database.

Configuration

Extend the native livekit configuration (.yaml file) with the following properties:

openvidu:
  analytics:
    enabled: true
    mongo_url: mongodb://localhost:27017/?replicaSet=rs0&readPreference=primaryPreferred
    interval: 10s
    expiration: 768h # 32 days
  • enabled: whether to enable the storage of events/stats or not.
  • mongo_url: URL of the MongoDB deployment where to store the events/stats. This is a Connection String.
  • interval: how often the events/stats batches must be sent. It is a time.Duration. Default is 10s.
  • expiration: the time to live of the events/stats in the storage destination. It is a time.Duration. Default is 768h (32 days).

Running locally

Just run the docker-compose.yml file to launch the required services (redis, mongo, and optional mongo-express):

docker compose up

Install dependencies:

go mod tidy

Run openvidu-livekit:

export LIVEKIT_CONFIG=$(cat ${PWD}/config.yaml)
go run ./...

Go to http://localhost:8081 to explore the MongoDB database with mongo-express (admin username and pass password).

Changing the destination of events/stats

It is fairly easy to adapt the code to store the events/stats in a different place, or even in multiple places at the same time. For example, file redis.go allows storing the events/stats in the very same Redis database used by the livekit deployment.

Any new implementation must comply with interface DatabaseClient:

type DatabaseClient interface {
	InitializeDatabase() error
	SendBatch()
	FixActiveEntities()
}

The InitializeDatabase method is used to prepare the destination database (for example creating indexes). The SendBatch method is used to send the events/stats in batches using the specific database client methods. The FixActiveEntities method is used to fix any active entities that are not actually active anymore.

Appart from complying with this interface, it is also required for any new implementation to have a reference to the LivekitHelper interface and the generic AnalyticsSender type:

type AnalyticsSender struct {
	eventsQueue    queue.Queue[*livekit.AnalyticsEvent]
	statsQueue     queue.Queue[*livekit.AnalyticsStat]
	databaseClient DatabaseClient
}

type BaseDatabaseClient struct {
	owner         *AnalyticsSender
	livekitHelper livekithelperinterface.LivekitHelper
}

This struct has two queues for storing events and stats separately in a FIFO manner, and a reference to the database client. The analytics service has a collection of AnalyticsSender, over which it iterates calling each SendBatch method.

With this in mind, this is the implementation of the MongoDB client:

type MongoDatabaseClient struct {
	BaseDatabaseClient
	client                *mongo.Client
}

func NewMongoDatabaseClient(conf *openviduconfig.AnalyticsConfig, livekithelper livekithelperinterface.LivekitHelper) (*MongoDatabaseClient, error) {
	mongoDatabaseClient := &MongoDatabaseClient{
		client: // {Golang MongoDB client},
	}

	sender := &AnalyticsSender{
		eventsQueue:    queue.NewSliceQueue[*livekit.AnalyticsEvent](),
		statsQueue:     queue.NewSliceQueue[*livekit.AnalyticsStat](),
		databaseClient: mongoDatabaseClient,
	}
	mongoDatabaseClient.owner = sender
	mongoDatabaseClient.livekitHelper = livekithelper

	return mongoDatabaseClient, nil
}

func (m *MongoDatabaseClient) InitializeDatabase() error {
	// Create indexes...
	return nil
}

func (m *MongoDatabaseClient) SendBatch() {
	// Send events using the golang MongoDB client...
	// Send stats using the golang MongoDB client...
}

func (m *MongoDatabaseClient) FixActiveEntities() {
	// Fix active entities...
}