This is a cookbook for managing RabbitMQ with Chef.
This cookbook targets Chef 14.0 and later.
5.x
release series targets RabbitMQ 3.8.x
releases.
For any series used, a supported Erlang version must be provisioned.
The cookbook generally targets the same Debian-based distribution versions that RabbitMQ core team supports:
- Recent and community supported versions of Ubuntu, e.g. 22.04 and later
- Debian 11 (Bullseye) and 12 (
- RHEL 8+
- CentOS Stream 9
- Fedora 35 or later
- Amazon Linux 2023
- Rocky Linux 9
- Alma Linux 9
Newer Debian, Ubuntu and RHEL/Fedora/CentOS Stream versions should work.
This cookbook depends on the Erlang cookbook and assumes that the user can configure it to provision a supported Erlang/OTP version.
Two more recipes are provided by this cookbook:rabbitmq::erlang_package
and rabbitmq::esl_erlang_package
.
The latter is an alias to the erlang::esl
recipe in the Erlang
cookbook.
The former uses Debian Erlang packages and zero dependency Erlang RPM package produced and published by Team RabbitMQ. Those packages provide the latest patch releases of Erlang/OTP.
Both options are covered below.
Provisioning RabbitMQ 3.9.x and 3.8.x (both out of support)
To provision RabbitMQ 3.9.x or 3.8.x, you must use version 5.8.5
of this cookbook or later.
Before provisioning a 3.9.x or 3.8.x release, please learn about the minimum required Erlang/OTP version and Erlang/OTP version recommendations.
Most distributions provide older versions, so Erlang must be provisioned either using RabbitMQ's zero dependency Erlang RPM, Debian Erlang packages, or from Erlang Solutions
rabbitmq::erlang_package
is a recipe that provisions latest Erlang packages from team RabbitMQ.
The packages support
- Debian 10 and 11
- Ubuntu 18.04 through 20.04
- RHEL 8
- CentOS Stream 8
- Fedora 32 or later
- Amazon Linux 2022
- Scientific Linux 7
The packages are cannot be installed alongside with other Erlang packages, for example, those from standard Debian repositories or Erlang Solutions.
To make sure that the Erlang cookbook is not used by rabbitmq::default
, rabbitmq::cluster
,
and other recipes, set node['rabbitmq']['erlang']['enabled']
to true
:
node['rabbitmq']['erlang']['enabled'] = true
By default rabbitmq::erlang_package
will install the latest Erlang version available.
To override package version, use node['rabbitmq']['erlang']['version']
:
# Debian
node['rabbitmq']['erlang']['version'] = '1:24.3.3-1'
# RPM
node['rabbitmq']['erlang']['version'] = '24.3.3'
On Ubuntu and Debian the distribution will be picked from node attributes.
Most of the time there is no need to override other attributes. Below is a list of defaults used on Ubuntu and Debian:
# RabbitMQ Erlang packages
deb_distro = value_for_platform(
'debian' => {
'default' => 'debian'
},
'ubuntu' => {
'default' => 'ubuntu'
}
)
default['rabbitmq']['erlang']['apt']['uri'] = "https://dl.cloudsmith.io/public/rabbitmq/rabbitmq-erlang/deb/#{deb_distro}"
default['rabbitmq']['erlang']['apt']['lsb_codename'] = node['lsb']['codename']
default['rabbitmq']['erlang']['apt']['components'] = ["main"]
default['rabbitmq']['erlang']['apt']['key'] = 'https://dl.cloudsmith.io/public/rabbitmq/rabbitmq-erlang/gpg.E495BB49CC4BBE5B.key'
default['rabbitmq']['erlang']['apt']['install_options'] = %w(--fix-missing)
On CentOS Stream 8, base Yum repository URL will be picked based on distribution versions. On Fedora, a suitable CentOS package will be used. Erlang package version is set the same way as for Debian (see above).
Below are the defaults used by the Yum repository (assuming RHEL or CentOS Stream 8):
# CentOS 8, RHEL 8, Fedora
default['rabbitmq']['erlang']['yum']['baseurl'] = 'https://dl.cloudsmith.io/public/rabbitmq/rabbitmq-erlang/rpm/el/8/$basearch'
default['rabbitmq']['erlang']['yum']['gpgkey'] = 'https://dl.cloudsmith.io/public/rabbitmq/rabbitmq-erlang/gpg.E495BB49CC4BBE5B.key'
default['rabbitmq']['erlang']['yum']['gpgcheck'] = true
default['rabbitmq']['erlang']['yum']['repo_gpgcheck'] = false
The Erlang cookbook will provision packages from Erlang Solutions
if node['erlang']['install_method']
is set to esl
.
Note that Erlang Solutions repositories can be behind the latest Erlang/OTP patch releases.
# will install the latest release, please
# consult with https://www.rabbitmq.com/which-erlang.html first
node['erlang']['install_method'] = "esl"
to provision a specific version, e.g. 24.3.3
:
node['erlang']['install_method'] = "esl"
# Ubuntu and Debian
# note the "1:" package epoch prefix
node['erlang']['esl']['version'] = "1:24.3.3"
node['erlang']['install_method'] = "esl"
# CentOS, RHEL, Fedora
node['erlang']['esl']['version'] = "24.3.3-1"
Set node['rabbitmq']['version']
to specify a version:
node['rabbitmq']['version'] = "3.9.14"
If you have node['rabbitmq']['deb_package_url']
or node['rabbitmq']['rpm_package_url']
overridden
from earlier versions, consider omitting those attributes. Otherwise see a section on download
location customization below.
RabbitMQ packages will be downloaded from Cloudsmith.
Installs rabbitmq-server
via direct download (from Bintray or GitHub, depending on the version) of
the installation package or using the distribution version. Depending on your distribution,
the provided version may be quite old so direct download is the default option.
Set the ['rabbitmq']['clustering']['enable']
attribute to true
, ['rabbitmq']['clustering']['cluster_disk_nodes']
array of node@host
strings that describe cluster members,
and a alphanumeric string for the erlang_cookie
.
To enable TLS for client connections, set the ssl
to true
and set the paths to your cacert, cert and key files.
node['rabbitmq']['ssl'] = true
# path to the CA bundle file
node['rabbitmq']['ssl_cacert'] = '/path/to/cacert.pem'
# path to the server certificate (pubic key) PEM file
node['rabbitmq']['ssl_cert'] = '/path/to/cert.pem'
# path to the server private key file
node['rabbitmq']['ssl_key'] = '/path/to/key.pem'
TCP connection listeners may be limited to a specific interface
using the node['rabbitmq']['tcp_listen_interface']
attribute.
Its TLS connection listener counterpart is node['rabbitmq']['ssl_listen_interface']
.
node['rabbitmq']['deb_package_url']
and node['rabbitmq']['rpm_package_url']
can be used
to override the package download location. They configure a prefix without a version.
Set them to a download location without a version if you want to provision from a custom
endpoint such as a local mirror.
The default
recipe will append a version suffix that matches RabbitMQ tag naming scheme.
For 3.7.x
or later, it is just the version (the value is used as is).
Lastly, a package name will be appended to form a full download URL. They rarely need
changing but can also be overridden using the node['rabbitmq']['deb_package']
and node['rabbitmq']['rpm_package']
attributes.
A full list of attributes related to TLS in RabbitMQ can be found in attributes/default.rb.
Default values and usage information of important attributes are shown below. More attributes are documented in metadata.rb.
The default username and password are guest
/guest
, with access limited to localhost connections:
['rabbitmq']['default_user'] = 'guest'
['rabbitmq']['default_pass'] = 'guest'
It is highly recommended that a different default user name is used with a reasonably long (say, 30-40 characters) generated password.
By default, the guest user can only connect from localhost. This prevents remote access for installations that use the well-known default credentials. It is highly recommended that remote access for the default user is not enabled but if security is of absolutely no importance in a certain environment, this can be done:
['rabbitmq']['loopback_users'] = []
Learn more in the RabbitMQ Access Control guide.
It is possible to to load a definitions (schema) file on node boot. Consult RabbitMQ's Definitions and Backup doc guides to learn more.
To configure definition loading, set the following attribute:
['rabbitmq']['management']['load_definitions'] = true
By default, the node will be configured to load a JSON at /etc/rabbitmq/load_definitions.json
;
however, you can define another path if you'd prefer using the following attribute:
['rabbitmq']['management']['definitions_file'] = '/path/to/your/definitions.json'
In order to use this functionality, you will need to provision a file referenced by the above attribute before you execute any recipes in the RabbitMQ cookbook (in other words, before the node starts). For example, this can be done using a remote file resource.
Installs the RabbitMQ management plugin.
To enable HTTPS for the management UI and HTTP API, set ['rabbitmq']['web_console_ssl']
attribute to true
.
The HTTPS port for the management UI can be configured by setting attribute ['rabbitmq']['web_console_ssl_port']
,
whose default value is 15671.
Enables any plugins listed in the node['rabbitmq']['enabled_plugins']
and disables any listed
in node['rabbitmq']['disabled_plugins']
attributes.
To enable the LDAP plugin, a few attributes have to be used in combination:
- Set
node['rabbitmq']['ldap']['enabled'] = true
- Enable
auth_backends
:node['rabbitmq']['auth_backends'] = 'rabbit_auth_backend_internal,rabbit_auth_backend_ldap'
- Enable the
rabbitmq_auth_backend_ldap
plugin - Configure LDAP servers and queries via the
node['rabbitmq']['ldap']['conf']
variable
# this is just an example
node['rabbitmq']['ldap']['conf'] = {
:servers => ["ldap-host1", "ldap-host2"],
:user_bind_pattern => "${username}@<domain>",
:dn_lookup_attribute => "sAMAccountName",
:dn_lookup_base => "DC=<CHANGEME>,DC=<CHANGEME>",
:port => <CHANGEME (number)>,
:log => <CHANGEME (boolean)>,
:vhost_access_query => '{constant, true}',
:topic_access_query => '{constant, true}',
:resource_access_query => '{constant, true}',
:tag_queries => "[{administrator, {constant, false}}]"
}
Enables any RabbitMQ users listed in the node['rabbitmq']['enabled_users']
and disables any listed in node['rabbitmq']['disabled_users']
attributes.
You can provide user credentials, the vhosts that they need to have access to and the permissions that should be allocated to each user.
node['rabbitmq']['enabled_users'] = [
{
:name => 'kitten',
:password => 'kitten',
:tag => 'leader',
:rights => [
{
:vhost => 'nova',
:conf => '.*',
:write => '.*',
:read => '.*'
}
]
}
]
Note that with this approach user credentials will be stored in the attribute file. Using encrypted data bags is therefore highly recommended.
Alternatively definitions export and import (see above) can be used. Definition files contain password hashes since clear text values are not stored.
Enables any virtual hosts listed in the node['rabbitmq']['virtualhosts']
and disables any listed in node['rabbitmq']['disabled_virtualhosts']
attributes.
Forms a cluster RabbitMQ of nodes.
It supports two clustering modes: auto or manual.
- Auto clustering: lists cluster members in the RabbitMQ config file. Those are taken from lists the nodes
node['rabbitmq']['clustering']['cluster_nodes']
. - Manual clustering : joins cluter members using
rabbitmqctl join_cluster
.
node['rabbitmq']['clustering']['enable']
: Default decision flag of clusteringnode['rabbitmq']['erlang_cookie']
: Same erlang cookie is required for the clusternode['rabbitmq']['clustering']['use_auto_clustering']
: Default is false. (manual clustering is default)node['rabbitmq']['clustering']['cluster_name']
: Name of cluster. default value is nil. In case of nil or '' is set forcluster_name
, first node name innode['rabbitmq']['clustering']['cluster_nodes']
attribute will be set for manual clustering. for the auto clustering, one of the node name will be set.node['rabbitmq']['clustering']['cluster_nodes']
: List of cluster nodes. it required node name and cluster node type. please refer to example in below.
Example
node['rabbitmq']['clustering']['enable'] = true
node['rabbitmq']['erlang_cookie'] = 'AnyAlphaNumericStringWillDo'
node['rabbitmq']['clustering']['cluster_partition_handling'] = 'pause_minority'
node['rabbitmq']['clustering']['use_auto_clustering'] = false
node['rabbitmq']['clustering']['cluster_name'] = 'qa_env'
node['rabbitmq']['clustering']['cluster_nodes'] = [
{
:name => 'rabbit@rabbit1'
},
{
:name => 'rabbit@rabbit2'
},
{
:name => 'rabbit@rabbit3'
}
]
Enables any policies listed in the node['rabbitmq']['policies']
and disables any listed in node['rabbitmq']['disabled_policies']
attributes.
See examples in attributes file.
Downloads, installs and enables pre-built community plugins binaries.
To specify a plugin, set the attribute node['rabbitmq']['community_plugins']['PLUGIN_NAME']
to '{DOWNLOAD_URL}'
.
There are several LWRPs for interacting with RabbitMQ and a few setting up Erlang repositories and package.
erlang_apt_repository_on_cloudsmith
sets up a Debian package repository from Cloudsmith.
It is a wrapper around the standard apt_repository
resource provider.
See also RabbitMQ Erlang Compatibility guide.
rabbitmq_erlang_apt_repository_on_cloudsmith 'rabbitmq_erlang_repo_on_cloudsmith' do
distribution node['lsb']['codename'] unless node['lsb'].nil?
action :add
end
erlang_apt_repository_on_cloudsmith
sets up an RPM package repository from Bintray.
It is a wrapper around the standard apt_repository
resource provider.
See also RabbitMQ Erlang Compatibility guide.
rabbitmq_erlang_yum_repository_on_cloudsmith 'rabbitmq_erlang' do
# for RHEL/CentOS 8+, Fedora. See https://www.rabbitmq.com/install-rpm.html.
baseurl 'https://dl.cloudsmith.io/public/rabbitmq/rabbitmq-erlang/rpm/el/8/$basearch'
action :add
end
Install the package. Here's an example for Debian-based systems:
rabbitmq_erlang_package_from_cloudsmith 'rabbitmq_erlang' do
# This package version assumes a Debian-based distribution.
version '1:24.3.3-1'
action :install
end
Here's another one for RPM-based ones:
rabbitmq_erlang_package_from_cloudsmith 'rabbitmq_erlang' do
version '24.3.3'
action :install
end
Enables or disables a rabbitmq plugin. Plugins are not supported for releases prior to 2.7.0.
:enable
enables aplugin
:disable
disables aplugin
rabbitmq_plugin "rabbitmq_stomp" do
action :enable
end
rabbitmq_plugin "rabbitmq_shovel" do
action :disable
end
Sets or clears a RabbitMQ policy.
:set
sets apolicy
:clear
clears apolicy
rabbitmq_policy "queue-length-limit" do
pattern "^limited\\.*"
definition ({"max-length" => "3000"})
priority 1
action :set
end
Adds and deletes users:
:add
adds auser
with apassword
:delete
deletes auser
:set_permissions
sets thepermissions
for auser
,vhost
is optional:clear_permissions
clears the permissions for auser
:set_tags
set the tags on a user:clear_tags
clear any tags on a user:change_password
set thepassword
for auser
rabbitmq_user "guest" do
action :delete
end
rabbitmq_user "nova" do
password "sekret"
action :add
end
rabbitmq_user "nova" do
vhost "/nova"
permissions ".* .* .*"
action :set_permissions
end
rabbitmq_user "rmq" do
vhost ["/", "/rmq", "/nova"]
permissions ".* .* .*"
action :set_permissions
end
rabbitmq_user "joe" do
tag "admin,lead"
action :set_tags
end
Adds and deletes virtual hosts.
:add
adds avhost
:delete
deletes avhost
rabbitmq_vhost "/nova" do
action :add
end
Forms a cluster and controls cluster name. This is an imperative version of classic config peer discovery in modern RabbitMQ versions.
-
:join
join in cluster as a manual clustering. node will join in first node of json string data. -
cluster nodes as JSON: all cluster nodes should be listed.
[
{
"name" : "rabbit@rabbit1",
"type" : "disc"
},
{
"name" : "rabbit@rabbit2",
"type": "disc"
},
{
"name" "rabbit@rabbit3",
"type" : "disc"
}
]
:set_cluster_name
set the cluster name
rabbitmq_cluster '[{"name":"rabbit@rabbit1", "type":"disc"},{"name":"rabbit@rabbit2", "type":"disc"},{"name":"rabbit@rabbit3", "type":"disc"}]' do
action :join
end
rabbitmq_cluster '[{"name":"rabbit@rabbit1","type":"disc"},{"name":"rabbit@rabbit2", ,"type":"disc"},{"name":"rabbit@rabbit3","type":"disc"}]' do
cluster_name 'seoul_tokyo_newyork'
action :set_cluster_name
end
rabbitmq_cluster '[{"name":"rabbit@rabbit1","type":"disc"},{"name":"rabbit@rabbit2", ,"type":"disc"},{"name":"rabbit@rabbit3","type":"disc"}]' do
action :change_cluster_node_type
end
This cookbook provides the primitives to remove a node from a cluster via helper functions but do not include these in any recipes. This is something that is potentially very dangerous and different deployments will have different needs and IF you decide you need this it should be implemented in your wrapper with EXTREME caution. There are 2 helper methods for 2 different scenario:
- removing self from cluster. This should likely only be considered for machines on a normal decommission. This is accomplished by using the helper fucntion
reset_current_node
. - removing another node from cluster. This should only be done once you are sure the machine is gone and won't come back. This can be accomplished via
remove_remote_node_from_cluster
.
For an already running cluster, these actions still require manual intervention:
- changing the shared cluster secret using the
:erlang_cookie
attribute - disabling clutering entirely by setting
:cluster
fromtrue
tofalse
- Author:: Benjamin Black
- Author:: Daniel DeLeo
- Author:: Matt Ray
- Author:: Seth Thomas
- Author:: JJ Asghar
- Author:: Team RabbitMQ
Copyright (c) 2009-2018, Chef Software, Inc.
Copyright (c) 2018-2021, VMware, Inc. or its affiliates.
Licensed under the Apache License, Version 2.0 (the "License");
you may not use this file except in compliance with the License.
You may obtain a copy of the License at
https://www.apache.org/licenses/LICENSE-2.0
Unless required by applicable law or agreed to in writing, software
distributed under the License is distributed on an "AS IS" BASIS,
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
See the License for the specific language governing permissions and
limitations under the License.