RabbitMQ projects use pull requests to discuss, collaborate on and accept code contributions. Pull requests is the primary place of discussing code changes.
The process is fairly standard:
- Present your idea to the RabbitMQ core team using GitHub Discussions or RabbitMQ community Discord server
- Fork the repository or repositories you plan on contributing to
- Run
git clean -xfffd && gmake clean && gmake distclean && gmake
to build all subprojects from scratch - Create a branch with a descriptive name
- Make your changes, run tests, ensure correct code formatting, commit with a descriptive message, push to your fork
- Submit pull requests with an explanation what has been changed and why
- Submit a filled out and signed Contributor Agreement if needed (see below)
- Be patient. We will get to your pull request eventually
Test suites of individual subprojects can be run from the subproject directory under
deps/*
. For example, for the core broker:
# Running all server suites in parallel will take between 30 and 40 minutes on reasonably
# recent multi-core machines. This is rarely necessary in development environments.
# Running individual test suites or groups of test suites can be enough.
#
# Before you start: this will terminate all running nodes, make processes and Common Test processes
killall -9 beam.smp; killall -9 erl; killall -9 make; killall -9 epmd; killall -9 erl_setup_child; killall -9 ct_run
# the core broker subproject
cd deps/rabbit
# cleans build artifacts
git clean -xfffd
gmake clean; gmake distclean
# builds the broker and all of its dependencies
gmake
# runs an integration test suite, tests/rabbit_fifo_SUITE with CT (Common Test)
gmake ct-rabbit_fifo
# runs an integration test suite, tests/quorum_queue_SUITE with CT (Common Test)
gmake ct-quorum_queue
# runs an integration test suite, tests/queue_parallel_SUITE with CT (Common Test)
gmake ct-queue_parallel
# runs a unit test suite tests/unit_log_management_SUITE with CT (Common Test)
gmake ct-unit_log_management
All ct-*
Make targets support a t=
argument which are transformed to -group
and -case
Common Test runner options.
# Runs a a group of tests named 'all_tests_with_prefix' in suite 'test/rabbit_mgmt_http_SUITE.erl'
gmake ct-rabbit_mgmt_http t="all_tests_with_prefix"
# Runs a test named 'users_test' in group 'all_tests_with_prefix' in suite 'test/rabbit_mgmt_http_SUITE.erl'
gmake ct-rabbit_mgmt_http t="all_tests_with_prefix:users_test"
# Runs a test named 'queues_test' in group 'all_tests_with_prefix' in suite 'test/rabbit_mgmt_http_SUITE.erl'
gmake ct-rabbit_mgmt_http t="all_tests_with_prefix:queues_test"
Set RABBITMQ_METADATA_STORE
to either khepri
or mnesia
to make the Common Test suites
use a specific schema data store (metadata store):
RABBITMQ_METADATA_STORE=khepri gmake ct-quorum_queue
Or, with Nu shell:
with-env {'RABBITMQ_METADATA_STORE': 'khepri'} { gmake ct-quorum_queue }
# Run from repository root.
# Starts a node with the management plugin enabled
gmake run-broker RABBITMQ_PLUGINS=rabbitmq_management
The nodes will be started in the background. They will use rabbit@{hostname}
for its name, so CLI will be able to contact
it without an explicit -n
(--node
) argument:
# Run from repository root.
./sbin/rabbitmq-diagnostics status
# Run from repository root.
# Starts a three node cluster with the management plugin enabled
gmake start-cluster NODES=3 RABBITMQ_PLUGINS=rabbitmq_management
The node will use rabbit-{n}@{hostname}
for names, so CLI must
be explicitly given explicit an -n
(--node
) argument in order to
contact one of the nodes:
rabbit-1
rabbit-2
rabbit-3
The names of the nodes can be looked up via
epmd -names
# Run from repository root.
# Makes CLI tools talk to node rabbit-2
sbin/rabbitmq-diagnostics cluster_status -n rabbit-2
# Run from repository root.
# Makes CLI tools talk to node rabbit-1
sbin/rabbitmq-diagnostics status -n rabbit-1
To stop a previously started cluster:
# Run from repository root.
# Stops a three node cluster started earlier
gmake stop-cluster NODES=3
When working on management UI code, besides starting the node with
# starts a node with the management plugin enabled
gmake run-broker RABBITMQ_PLUGINS=rabbitmq_management
(or any other set of plugins), it is highly recommended to use BrowserSync to shorten the edit/feedback cycle for JS files, CSS, and so on.
First, install BrowserSync using NPM:
npm install -g browser-sync
Assuming a node running locally with HTTP API on port 15672
, start
a BrowserSync proxy like so:
cd deps/rabbitmq_management/priv/www
browser-sync start --proxy localhost:15672 --serverStatic . --files .
BrowserSync will automatically open a browser window for you to use. The window will automatically refresh when one of the static (templates, JS, CSS) files change.
All HTTP requests that BrowserSync does not know how to handle will be proxied to
the HTTP API at localhost:15672
.
The RabbitMQ CLI uses the standard Elixir code formatter. To ensure correct code formatting of the CLI:
cd deps/rabbitmq_cli
mix format
Running make
will validate the CLI formatting and issue any necessary warnings. Alternatively, run the format checker in the deps/rabbitmq_cli
directory:
mix format --check-formatted
See CODE_OF_CONDUCT.md.
If you want to contribute a non-trivial change, please submit a signed copy of our Contributor Agreement around the time you submit your pull request. This will make it much easier (in some cases, possible) for the RabbitMQ team at Pivotal to merge your contribution.
If something isn't clear, feel free to ask on GitHub Discussions and community Discord server.