GarageMQ

GarageMQ is a message broker that implement the Advanced Message Queuing Protocol (AMQP). Compatible with any AMQP or RabbitMQ clients (tested streadway/amqp and php-amqp lib)

Table of Contents

• Goals of this project
• Demo
• Installation and Building
  • Docker
  • Go get
  • Execution flags
  • Default config params
• Performance tests
  • Persistent messages
  • Transient messages
• Internals
  • Backend for durable entities
  • QOS
  • Admin server
• TODO
• Contribution

Goals of this project

• Have fun and learn a lot
• Implement AMQP message broker in Go
• Make protocol compatible with RabbitMQ and standard AMQP 0-9-1.

Demo

Simple demo server on Digital Ocean, 2 GB Memory / 25 GB Disk / FRA1 - Ubuntu Docker 17.12.0~ce on 16.04

Server	Port	Admin port	Login	Password	Virtual Host
46.101.117.78	5672	15672	guest	guest	/

• AdminServer - http://46.101.117.78:15672/
• Connect uri - amqp://guest:guest@46.101.117.78:5672

Installation and Building

Docker

The quick way to start with GarageMQ is by using docker. You can build it by your own or pull from docker-hub

docker pull amplitudo/garagemq
docker run --name garagemq -p 5672:5672 -p 15672:15672 amplitudo/garagemq

or

go get -u github.com/valinurovam/garagemq/...
cd $GOPATH/src/github.com/valinurovam/garagemq
docker build -t garagemq .
docker run --name garagemq -p 5672:5672 -p 15672:15672 garagemq

Go get

You can also use go get: go get -u github.com/valinurovam/garagemq/...

go get -u github.com/valinurovam/garagemq/...
cd $GOPATH/src/github.com/valinurovam/garagemq
make build.all && make run

Execution flags

Flag	Default	Description	ENV
--config	default config	Config path	GMQ_CONFIG
--log-file	stdout	Log file path or stdout, stderr	GMQ_LOG_FILE
--log-level	info	Logger level	GMQ_LOG_LEVEL
--hprof	false	Enable or disable hprof profiler	GMQ_HPROF
--hprof-host	0.0.0.0	Profiler host	GMQ_HPROF_HOST
--hprof-port	8080	Profiler port	GMQ_HPROF_PORT

Default config params

# Proto name to implement (amqp-rabbit or amqp-0-9-1)
proto: amqp-rabbit
# User list
users:
  - username: guest
    password: 084e0343a0486ff05530df6c705c8bb4 # guest md5
# Server TCP settings
tcp:
  ip: 0.0.0.0
  port: 5672
  nodelay: false
  readBufSize: 196608
  writeBufSize: 196608
# Admin-server settings
admin:
  ip: 0.0.0.0
  port: 15672
queue:
  shardSize: 8192
  maxMessagesInRam: 131072
# DB settings
db:
  # default path 
  defaultPath: db
  # backend engine (badger or buntdb) 
  engine: badger
# Default virtual host path  
vhost:
  defaultPath: /
# Security check rule (md5 or bcrypt)
security:
  passwordCheck: md5
connection:
  channelsMax: 4096
  frameMaxSize: 65536

Performance tests

Performance tests with load testing tool https://github.com/rabbitmq/rabbitmq-perf-test on test-machine:

MacBook Pro (15-inch, 2016)
Processor 2,6 GHz Intel Core i7
Memory 16 GB 2133 MHz LPDDR3

Persistent messages

./bin/runjava com.rabbitmq.perf.PerfTest --exchange test -uri amqp://guest:guest@localhost:5672 --queue test --consumers 10 --producers 5 --qos 100 -flag persistent
...
...
id: test-235131-686, sending rate avg: 53577 msg/s
id: test-235131-686, receiving rate avg: 51941 msg/s

Transient messages

./bin/runjava com.rabbitmq.perf.PerfTest --exchange test -uri amqp://guest:guest@localhost:5672 --queue test --consumers 10 --producers 5 --qos 100
...
...
id: test-235231-085, sending rate avg: 71247 msg/s
id: test-235231-085, receiving rate avg: 69009 msg/s

Internals

Backend for durable entities

Database backend is changeable through config db.engine

db:
  defaultPath: db
  engine: badger

db:
  defaultPath: db
  engine: buntdb

• Badger https://github.com/dgraph-io/badger
• BuntDB https://github.com/tidwall/buntdb

QOS

basic.qos method implemented for standard AMQP and RabbitMQ mode. It means that by default qos applies for connection(global=true) or channel(global=false). RabbitMQ Qos means for channel(global=true) or each new consumer(global=false).

Admin server

The administration server is available at standard :15672 port and is read only mode at the moment. Main page above, and more screenshots at /readme folder

TODO

• Optimize binds
• Replication and clusterization
• Own backend for durable entities and persistent messages
• Migrate to message reference counting

Contribution

Contribution of any kind is always welcome and appreciated. Contribution Guidelines in WIP