RabbitMQ Stream tutorial - "Hello World!"
Introduction
Prerequisites
This tutorial assumes RabbitMQ is installed, running on
localhost
and the stream plugin enabled.
The standard stream port is 5552. In case you
use a different host, port or credentials, connections settings would require
adjusting.
Using docker
If you don't have RabbitMQ installed, you can run it in a Docker container:
docker run -it --rm --name rabbitmq -p 5552:5552 -p 15672:15672 -p 5672:5672 \
-e RABBITMQ_SERVER_ADDITIONAL_ERL_ARGS='-rabbitmq_stream advertised_host localhost' \
rabbitmq:3.13
wait for the server to start and then enable the stream and stream management plugins:
docker exec rabbitmq rabbitmq-plugins enable rabbitmq_stream rabbitmq_stream_management
Where to get help
If you're having trouble going through this tutorial you can contact us through the mailing list or discord community server.
RabbitMQ Streams was introduced in RabbitMQ 3.9. More information is available here.
"Hello World"
(using the Java Stream Client)
In this part of the tutorial we'll write two programs in Java; a producer that sends a single message, and a consumer that receives messages and prints them out. We'll gloss over some of the details in the Java client API, concentrating on this very simple thing just to get started. It's the "Hello World" of RabbitMQ Streams.
The Java stream client library
RabbitMQ speaks multiple protocols. This tutorial uses RabbitMQ stream protocol which is a dedicated protocol for RabbitMQ streams. There are a number of clients for RabbitMQ in many different languages, see the stream client libraries for each language. We'll use the Java stream client provided by RabbitMQ.
RabbitMQ Java client 0.16.0 and later versions are distributed via Maven Repository.
This tutorial assumes you are using powershell on Windows. On MacOS and Linux nearly any shell will work.
Setup
This tutorial requires java
to be in PATH
. To verify it, run:
java --help
This tutorial will use Maven to manage dependencies and build the project. It is not necessary to install Maven as the tutorial uses the Maven Wrapper. An executable version of this tutorial can be found in the RabbitMQ tutorials repository.
Next, let's verify if Maven works properly:
./mvnw --version
Next, create a pom.xml
file with the RabbitMQ Stream Java client as a dependency:
<dependency>
<groupId>com.rabbitmq</groupId>
<artifactId>stream-client</artifactId>
<version>0.16.0</version>
</dependency>
Sending
Next, let's create two files for the message producer (sender) and the message consumer (receiver) part of this tutorial.
They will be called Send.java
and
Receiver.java
, respectively.
The producer will connect to RabbitMQ, send a single message, then exit. The consumer will consume and print it to standard output.
At the top of Send.java
, import a few key classes this tutorial uses:
import com.rabbitmq.stream.*;
import java.io.IOException;
With these classes imported, an Environment
now can be instantiated:
Environment environment = Environment.builder().build();
The entry point of the stream Java client is the Environment
.
It is used for configuration of RabbitMQ stream publishers, stream consumers,
and streams themselves.
It abstracts away a TCP or TLS socket connection, and takes care of protocol version negotiation and authentication and so on for us.
This tutorial assumes that stream publisher and consumer connect to
a RabbitMQ node running locally, that is, on localhost. To connect to a node on a different
machine, simply specify target hostname or IP address using the builder
returned by Environment.builder()
.
Next, let's create a producer.
The producer will also declare a stream it will publish messages to and then publish a message:
String stream = "hello-java-stream";
environment.streamCreator().stream(stream).maxLengthBytes(ByteCapacity.GB(5)).create();
Producer producer = environment.producerBuilder().stream(stream).build();
producer.send(producer.messageBuilder().addData("Hello, World!".getBytes()).build(), null);
System.out.println(" [x] 'Hello, World!' message sent");
The stream declaration operation is idempotent: the stream will only be created if it doesn't exist already.
A stream is an append-only log abstraction that allows for repeated consumption of messages until they expire. It is a good practice to always define the retention policy. In the example above, the stream is limited to be 5 GiB in size.
The message content is a byte array. Applications can encode the data they need to transfer using any appropriate format such as JSON, MessagePack, and so on.
When the code above finishes running, the producer connection and stream-system connection will be closed. That's it for our producer.
Each time the producer is run, it will send a single message to the server and the message will be appended to the stream.
The complete Send.java
file can
be found on GitHub.
Sending doesn't work!
If this is your first time using RabbitMQ and you don't see the "Sent" message then you may be left scratching your head wondering what could be wrong. Maybe the broker was started without enough free disk space (by default it needs at least 50 MB free) and is therefore refusing to accept messages. Check the broker log file to see if there is a resource alarm logged and reduce the free disk space threshold if necessary. The Configuration guide will show you how to set
disk_free_limit
.Another reason may be that the program exits before the message makes it to the broker. Sending is asynchronous in some client libraries: the function returns immediately but the message is enqueued in the IO layer before going over the wire. The sending program asks the user to press a key to finish the process: the message has plenty of time to reach the broker. The stream protocol provides a confirm mechanism to make sure the broker receives outbound messages, but this tutorial does not use this mechanism for simplicity's sake.
Receiving
The other part of this tutorial, the consumer, will connect to a RabbitMQ node and wait for messages to be pushed to it. Unlike the producer, which in this tutorial publishes a single message and stops, the consumer will be running continuously, consume the messages RabbitMQ will push to it, and print the received payloads out.
Similarly to Send.java
, Receive.java
will need some classes to be imported first:
import com.rabbitmq.stream.ByteCapacity;
import com.rabbitmq.stream.Consumer;
import com.rabbitmq.stream.Environment;
import com.rabbitmq.stream.OffsetSpecification;
When it comes to the initial setup, the consumer part is very similar the producer one; we use the default connection settings and declare the stream from which the consumer will consume.
Note that the stream name must match that used by the producer.
Environment environment = Environment.builder().build();
String stream = "hello-java-stream";
environment.streamCreator().stream(stream).maxLengthBytes(ByteCapacity.GB(5)).create();
Note that the consumer part also declares the stream. This is to allow either part to be started first, be it the producer or the consumer.
The Consumer
class is used to instantiate a stream consumer and environment.consumerBuilder()
provides a builder object that configures it.
Finally, the .messageHandler
method accepts a handler for delivered messages.
The offset
parameter defines the starting point of the consumer.
In this case, the consumer starts from the very first message available in the stream.
Consumer consumer = environment.consumerBuilder()
.stream(stream)
.offset(OffsetSpecification.first())
.messageHandler((unused, message) -> {
System.out.println("Received message: " + new String(message.getBodyAsBinary()));
}).build();
The complete Receive.java
file
can be found on GitHub.
Putting It All Together
In order to run both examples, open two terminal (shell) tabs.
Both parts of this tutorial can be run in any order, as they both declare the stream. Let's run the consumer first so that when the first publisher is started, the consumer will print it:
./mvnw -q compile exec:java '-Dexec.mainClass=Receive'
Then run the producer:
./mvnw -q compile exec:java '-Dexec.mainClass=Send'
The consumer will print the message it gets from the publisher via RabbitMQ. The consumer will keep running, waiting for new deliveries. Try re-running the publisher several times to observe that.
Streams are different from queues in that they are append-only logs of messages that can be consumed repeatedly. When multiple consumers consume from a stream, they will start from the first available message.