UCloud supports event streams, an event stream provides a message pipe. Producers send messages into the pipe and consumers read and process the messages. We support two types of event streams: EventStreams and BroadcastingStreams. UCloud implements both types using Redis.

Ordinary Event Streams

An ordinary EventStream provides a way for services to load-balance events between instances of a given micro-service. As shown in the figure, the example system contains to micro-services A and B. Both of these micro-services are running in three instances of the same code. Both micro-services have subscribed to a the same EventStream. When the producer starts producing messages, those messages are load-balanced across the individual instances of a micro-services. This means that both service A and service B receives all the messages but the individual messages are load balanced between the instances of a micro-service.

A consumer does not need to be live when the message is sent for it to be received. Instead, the messages are kept in a persistent list which lives for some time. This list is pruned every once in a while.

Example: Consuming a message

object JobEvents : EventStreamContainer() {
    val events = stream<JobEvent>("app-kubernetes-job-events", { it.jobName })
suspend fun initializeConsumer() {
    micro.eventStreamService.subscribe(, EventConsumer.Immediate { ev ->
        println("received $ev")

Example: Consuming a batch of messages

suspend fun initializeConsumer() {
        EventConsumer.Batched(maxLatency = 500, maxBatchSize = 1000) { batch ->
            println("received $batch")

Example: Producing a message

val eventProducer = micro.eventStreamService.createProducer(

Broadcasting Streams

The BroadcastingStream works differently from the EventStream. In this case all consumers receive every message which is produced.

Only the consumers which are live will receive messages.

Example: Broadcasting a message

val broadcastingStream = RedisBroadcastingStream(micro.redisConnectionManager)

Example: Subscribing to a stream

broadcastStream.subscribe( { (id) ->