Dump messages from a RabbitMQ queue to files, without affecting the queue.

Precompiled binary packages can be found on the releases page.

If you have Go installed, you can install rabbitmq-dump-queue from source by running:

go get github.com/dubek/rabbitmq-dump-queue

The rabbitmq-dump-queue executable will be created in the $GOPATH/bin directory.

To dump the first 50 messages of queue incoming_1 to /tmp:

rabbitmq-dump-queue -uri="amqp://user:[email protected]:5672/" -queue=incoming_1 -max-messages=50 -output-dir=/tmp

This will create the files /tmp/msg-0000, /tmp/msg-0001, and so on.

The output filenames are printed one per line to the standard output; this allows piping the output of rabbitmq-dump-queue to xargs or similar utilities in order to perform further processing on each message (e.g. decompressing, decoding, etc.).

To include the AMQP headers and properties in the output, add the -full option to the command-line. This will create the following files:

/tmp/msg-0000
/tmp/msg-0000-headers+properties.json
/tmp/msg-0001
/tmp/msg-0001-headers+properties.json
...

The JSON files have the following structure:

{
  "headers": {
    "x-my-private-header": "my-value"
  },
  "properties": {
    "correlation_id": "XYZ-9876",
    "delivery_mode": 0,
    "priority": 5
  }
}

By default, it will not acknowledge messages, so they will be requeued. Acknowledging messages using the -ack=true switch will remove them from the queue, allowing the user to process new messages (see implementation details).

rabbitmq-dump-queue -uri="amqp://user:[email protected]:5672/" -queue=incoming_1 -max-messages=50 -output-dir=/tmp -ack=true

Running rabbitmq-dump-queue -help will list the available command-line options.

In order to fetch messages from the queue and later return them in the original order, rabbitmq-dump-queue uses a standard AMQP basic.get API call without automatic acknowledgements, and it doesn't manually acknowledge the received messages. Thus, when the AMQP connection is closed (after all the messages were received and written to files), RabbitMQ returns all the un-acked messages (all the messages) back to the queue in their original order.

This means that during the time rabbitmq-dump-queue receives and saves the messages, the messages are not visible to other consumers of the queue. This duration is usually very short (unless you're downloading a lot of messages), but make sure your system can handle such a situation (or shut down other consumers of the queue during the time you use this tool).

Note that the same approach is used by RabbitMQ's management HTTP API (the /api/queues/{vhost}/{queue}/get endpoint with requeue=true).

To run the automated tests, have a RabbitMQ server listen on 127.0.0.1:5672 with user guest and password guest. This can be easily achieved with Docker:

docker run --name test-rabbitmq -d -p 5672:5672 rabbitmq:3-management

Then run:

go build .
go test -v .

To stop the RabbitMQ server container, run:

docker stop test-rabbitmq
docker rm test-rabbitmq

Github pull requests and issues are welcome.

rabbitmq-dump-queue is under the MIT License. See the LICENSE file for details.