Rabbit saving extra bytes for AMQP 1.0 messages - rabbitmq

Question

I have an environment where some AMQP 1.0 and some AMQP 0.9.1 clients need to write to/read from a RabbitMQ queue. I enabled the AMQP 1.0 rabbit plugin and it is working, but I get extra bytes in body for each AMQP 1.0 message.
I'm sending messages through AMQP1.0 by using rhea (typescript):
const connection: Connection = new Connection (
{
host: 'localhost',
port: 5672,
id: 'my_id',
reconnect: true
}
);
const senderName = "sender01";
const senderOptions: SenderOptions = {
name: senderName,
target: {
address: "target.queue"
},
onError: (context: EventContext) => {},
onSessionError: (context: EventContext) => {}
};
await connection.open();
const sender: Sender = await connection.createSender(senderOptions);
sender.send({
body: JSON.stringify({"one": "two", "three": "four"}),
content_encoding: 'UTF-8',
content_type: 'application/json'
});
console.log("sent");
await sender.close();
await connection.close();
console.log("connection closed");
This example works but this is what is stored in the queue:
The base64 encoded message is AFN3oRx7Im9uZSI6InR3byIsInRocmVlIjoiZm91ciJ9, which after decoding becomes:
Sw{"one":"two","three":"four"}
There is an additional Sw which I didn't send.
I tried setting up a java client with the official RabbitMQ library (which talks AMQP 0.9.1) to see if those extra bytes were sent to clients:
ConnectionFactory factory = new ConnectionFactory();
factory.setHost("localhost");
Connection connection = factory.newConnection();
Channel channel = connection.createChannel();
channel.basicConsume(
"target.queue",
true,
(consumerTag, delivery) -> {
String message = new String(delivery.getBody(), "UTF-8");
System.out.println(" [x] Received '" + message + "'");
},
ignored -> {}
);
This is the output:
[x] Received ' Sw�{"one":"two","three":"four"}'
The weird thing is that if I try consuming the exact same message with an AMQP 1.0 client, those extra bytes don't appear in the received message body, extra bytes appear only when publishing with AMQP 1.0 and subscribing with AMQP 0.9.1.
Why is that? Is there any way to avoid extra bytes when using both AMQP versions?
UPDATE
I also tried with SwiftMQ:
int nMsgs = 100;
int qos = QoS.AT_MOST_ONCE;
AMQPContext ctx = new AMQPContext(AMQPContext.CLIENT);
String host = "localhost";
int port = 5672;
String queue = "target.queue";
try {
Connection connection = new Connection(ctx, host, port, false);
connection.setContainerId(UUID.randomUUID().toString());
connection.setIdleTimeout(-1);
connection.setMaxFrameSize(1024 * 4);
connection.setExceptionListener(Exception::printStackTrace);
connection.connect();
{
Session session = connection.createSession(10, 10);
Producer p = session.createProducer(queue, qos);
for (int i = 0; i < nMsgs; i++) {
AMQPMessage msg = new AMQPMessage();
System.out.println("Sending " + i);
msg.setAmqpValue(new AmqpValue(new AMQPString("{\"one\":\"two\",\"three\":\"four\"}")));
p.send(msg);
}
p.close();
session.close();
}
connection.close();
} catch (Exception e) {
e.printStackTrace();
}
The issue is still there, but the first bytes changed, now I get:
[x] Received '□�□□□□□□□w�{"one":"two","three":"four"}'

Answer 1

Please see this response which clarifies how to encode the payload and avoid these extra bytes:
if AMQP 1.0 client sends message to 0-9-1 client and encodes her payload as binary in a "data section" (i.e. not in amqp-sequence section, not in amqp-value section) the 0-9-1 client shall get the the complete payload without any extra bytes
NOTE: the RabbitMQ team monitors the rabbitmq-users mailing list and only sometimes answers questions on StackOverflow.