Build Status Build Status


A message queue server that implements the AMQP 0-9-1 protocol. Written in Crystal.

Aims to be very fast, has low RAM requirements, handles very long queues, many connections, and requires minimal configuration.



From source

Begin with installing Crystal. Refer to Crystal's installation documentation on how to install Crystal.

Clone the git repository and build the project.

git clone [email protected]:cloudamqp/lavinmq.git
cd lavinmq
sudo make install # optional

Now, LavinMQ is ready to be used. You can check the version with:

lavinmq -v


curl -fsSL | gpg --dearmor | sudo tee /usr/share/keyrings/lavinmq.gpg > /dev/null
. /etc/os-release
echo "deb [signed-by=/usr/share/keyrings/lavinmq.gpg]$ID $VERSION_CODENAME main" | sudo tee /etc/apt/sources.list.d/lavinmq.list
sudo apt-get update
sudo apt-get install lavinmq

If you need to install a specific version of LavinMQ, do so using the following command: sudo apt install lavinmq=<version>. This works for both upgrades and downgrades.


sudo tee /etc/yum.repos.d/lavinmq.repo << 'EOF'
sudo dnf install lavinmq


LavinMQ only requires one argument, and it's a path to a data directory.

Run LavinMQ with: lavinmq -D /var/lib/lavinmq

More configuration options can be viewed with -h, and you can specify a configuration file too, see extras/lavinmq.ini for an example.


Docker images are published to Docker Hub. Fetch and run the latest version with:

docker run --rm -it -p 5672:5672 -p 15672:15672 -v /tmp/amqp:/var/lib/lavinmq cloudamqp/lavinmq

You are then able to visit the management UI at http://localhost:15672 and start publishing/consuming messages to amqp://guest:guest@localhost.


In Linux, perf is the tool of choice when tracing and measuring performance.

To see which syscalls that are made use:

perf trace -p $(pidof lavinmq)

To get a live analysis of the mostly called functions, run:

perf top -p $(pidof lavinmq)

A more detailed tutorial on perf is available here.

In OS X the app, Instruments that's bundled with Xcode can be used for tracing.

Memory garbage collection can be diagnosed with boehm-gc environment variables.


Kindly read our contributing guide

LavinMQ with various plattforms

All AMQP client libraries work with LavinMQ and there are AMQP client libraries for almost every platform on the market. Here are guides for a couple of common plattforms.

  1. Ruby
  2. Node.js
  3. Java
  4. Python
  5. PHP
  6. Crystal


A single m6g.large EC2 instance, with a GP3 EBS drive (XFS formatted), can sustain about 700.000 messages/s (16 byte msg body, single queue, single producer, single consumer). A single producer can push 1.600.000 msgs/s and if there's no producers auto-ack consumers can receive 1.200.000 msgs/s.

Enqueueing 100M messages only uses 25 MB RAM. 8000 connection uses only about 400 MB RAM. Declaring 100.000 queues uses about 100 MB RAM. About 1.600 bindings per second can be made to non-durable queues, and about 1000 bindings/second to durable queues.


LavinMQ is written in Crystal, a modern language built on the LLVM, with a Ruby-like syntax. It uses an event loop library for IO, is garbage collected, adopts a CSP-like concurrency model and compiles down to a single binary. You can liken it to Go, but with a nicer syntax.

Instead of trying to cache messages in RAM, we write all messages as fast as we can to disk and let the OS cache do the caching.

Each queues is backed by a message store on disk, which is just a series of files (segments), by default 8MB each. Message segments are memory-mapped files allocated using the mmap syscall. However, to prevent unnecessary memory usage, we unmap these files and free up the allocated memory when they are not in use. When a file needs to be written or read, we re-map it and use only the memory needed for that specific segment. Each incoming message is appended to the last segment, prefixed with a timestamp, its exchange name, routing key and message headers.

When a message is being consumed it reads sequentially from the segments. Each acknowledged (or rejected) message position in the segment is written to an "ack" file (per segment). If a message is requeued its position is added to a in memory queue. On boot all acked message positions are read from the "ack" files and then when deliviering messages skip those when reading sequentially from the message segments. Segments are deleted when all message in them are acknowledged.

Declarations of queues, exchanges and bindings are written to a definitions file (if the target is durable), encoded as the AMQP frame they came in as. Periodically this file is compacted/garbage-collected by writing only the current in-memory state to the file (getting rid of all delete events). This file is read on boot to restore all definitions.

All non-AMQP objects like users, vhosts, policies, etc. are stored in JSON files. Most often these type of objects does not have a high turnover rate, so we believe that JSON in this case makes it easy for operators to modify things when the server is not running, if ever needed.

In the data directory we store users.json and vhosts.json as mentioned earlier, and each vhost has a directory in which we store definitions.amqp (encoded as AMQP frames), policies.json and the messages named such as msgs.0000000124. Each vhost directory is named after the sha1 hash of its real name. The same goes for the queue directories in the vhost directory. The queue directories only has two files, ack and enq, also described earlier.


Here is an architectural description of the different flows in the server.


Client#read_loop reads from the socket, it calls Channel#start_publish for the Basic.Publish frame and Channel#add_content for Body frames. When all content has been received (and appended to an IO::Memory object) it calls VHost#publish with a Message struct. VHost#publish finds all matching queues, writes the message to the message store and then calls Queue#publish with the segment position. Queue#publish writes to the message store.


When Client#read_loop receives a Basic.Consume frame it will create a Consumer class and add it to the queue's list of consumers. Each consumer has a deliver_loop fiber that will be notified by an internal Channel when new messages are available in the queue.

Getting help

For questions or suggestions:


Currently missing but planned features

Known differences to other AMQP servers

There are a few edge-cases that are handled a bit differently in LavinMQ compared to other AMQP servers.


LavinMQ can be fully clustered with multiple other LavinMQ nodes. One node is always the leader and the others stream all changes in real-time. Failover happens instantly when the leader is unavailable.

etcd is used for leader election and maintaining the In-Sync-Replica (ISR) set. LavinMQ then uses a custom replication protocol between the nodes. When a follower disconnects it will fall out of the ISR set, and will then not be eligible to be a new leader.

Clustering configuration

Enable clustering with the following config:

enabled = true
bind = ::
port = 5679
advertised_uri = tcp://my-ip:5679
etcd_endpoints = localhost:2379

or start LavinMQ with:

lavinmq --data-dir /var/lib/lavinmq --clustering --clustering-bind :: --clustering-advertised-uri=tcp://my-ip:5679

Stream queues

Stream queues are like append-only logs and can be consumed multiple times. Each consumer can start to read from anywhere in the queue (using the x-stream-offset consumer argument) over and over again. Stream queues are different from normal queues in that messages are not deleted (see #retention) when a consumer acknowledge them.


Messages are only deleted when max-length, max-length-bytes or max-age are applied, either as queue arguments or as policies. The limits are checked only when new messages are published to the queue, and only act on whole segments (which by default are 8MiB), so the limits aren't necessarily exact. So even if a max-age limit is set, but no messages are published to the queue, messages might still be available in the stream queue that is way older that the limit specified.



The software is licensed under the Apache License 2.0.

Copyright 2018-2024 84codes AB

LavinMQ is a trademark of 84codes AB