Federation Plugin
Overview
The high-level goal of the Federation plugin is to transmit messages between brokers without requiring clustering. This is useful for a number of reasons.
Loose Coupling of Nodes or Clusters
The federation plugin can transmit messages between brokers (or clusters) in different administrative domains:
- they may have different users and virtual hosts;
- they may run on different versions of RabbitMQ and Erlang
WAN friendliness
The federation plugin uses AMQP 0-9-1 to communicate between brokers, and is designed to tolerate intermittent connectivity.
Specificity
A broker can contain federated and local-only components to best fit the desired architecture of the system.
Scalability with Growing Connected Node Count
Federation does not require O(n2) connections between N brokers (although this is the easiest way to set things up), which should mean it scales better.
What Does It Do?
The federation plugin allows you to make exchanges and queues federated. A federated exchange or queue can receive messages from one or more upstreams (remote exchanges and queues on other brokers). A federated exchange can route messages published upstream to a local queue. A federated queue lets a local consumer receive messages from an upstream queue.
Federation links connect to upstreams using RabbitMQ Erlang client. Therefore they can connect to a specific vhost, use TLS, use multiple authentication mechanisms.
For more details, see the documentation on federated exchanges and federated queues.
How is Federation Set Up?
To use federation, one needs to configure two things
- One or more upstreams that define federation connections to other nodes. This can be done via runtime parameters or the federation management plugin which adds a federation management tab to the management UI.
- One or more policies that match exchanges/queues and makes them federated.
Getting Started
The federation plugin is included in the RabbitMQ distribution. To enable it, use rabbitmq-plugins:
rabbitmq-plugins enable rabbitmq_federation
When using the management plugin, you will also want to
enable rabbitmq_federation_management
:
rabbitmq-plugins enable rabbitmq_federation_management
When using a federation in a cluster, all the nodes of the cluster should have the federation plugin enabled.
Information about federation upstreams is stored in the RabbitMQ database, along with users, permissions, queues, etc. There are three levels of configuration involved in federation:
- Upstreams: each upstream defines a remote connection endpoint.
- Upstream sets: each upstream set groups together a set of upstreams to use for federation.
- Policies: each policy selects a set of exchanges, queues or both, and applies a single upstream or an upstream set to those objects.
In practice, for simple use cases you can almost ignore the
existence of upstream sets, since there is an implicitly-defined upstream set called all
to which all upstreams are added.
Upstreams and upstream sets are both instances of runtime parameters. Like exchanges and queues, each virtual host has its own distinct set of parameters and policies. For more generic information on parameters and policies, see the guide on parameters and policies. For full details on the parameters used by federation, see the federation reference.
Parameters and policies can be set in three ways - either with
an invocation of rabbitmqctl
, a call to the
management HTTP API, or (usually) through the web UI presented
by rabbitmq_federation_management
. The HTTP API
does not present all possibilities - in particular, it does not support management of upstream sets.
A Basic Example
Here we will federate all the built-in exchanges except for the default exchange, with a single upstream. The upstream will be defined to buffer messages when disconnected for up to one hour (3600000ms).
First let's define an upstream:
rabbitmqctl |
|
---|---|
rabbitmqctl.bat (Windows) |
|
HTTP API |
|
Web UI | Navigate to |
Then define a policy that will match built-in exchanges and use this upstream:
rabbitmqctl |
|
---|---|
rabbitmqctl (Windows) |
|
HTTP API |
|
Management UI | Navigate to |
We tell the policy to federate all exchanges whose names
begin with "amq." (i.e. all the built in exchanges except
for the default exchange) with (implicit) low priority, and
to federate them using the implicitly created upstream set
"all", which includes our newly-created upstream. Any other
matching policy with a priority greater than 0 will take
precedence over this policy. Keep in mind that federate-me
is just a name we used for this example, you can use any
string you want there.
The built in exchanges should now be federated because they are matched by the policy. You can check that the policy has applied to the exchanges by checking the exchanges list in management or with:
rabbitmqctl list_exchanges name policy | grep federate-me
And you can check that federation links for each exchange have come up with Admin
> Federation Status
> Running Links
or with:
# This command will be available only if federation plugin is enabled
rabbitmqctl federation_status
In general there will be one federation link for each upstream that is applied to an exchange. So for example with three exchanges and two upstreams for each there will be six links.
For simple use this should be all you need - you will probably want to look at the AMQP URI reference.
The federation reference contains more details on upstream parameters and upstream sets.
Federation Connection (Link) Failures
Inter-node connections used by Federation are based on AMQP 0-9-1 connections. Federation links can be treated as special kind of clients by operators.
Should a link fail, e.g. due to a network interruption, it will attempt to re-connect. Reconnection period is a configurable value that's defined in upstream definition. See federation reference for more details on setting up upstreams and upstream sets.
Links generally try to recover ad infinitum but there are scenarios when they give up:
- Failure rate is too high (max tolerated rate depends on
upstream's
reconnect-delay
but is generally a failure every few seconds by default). - Link no longer can locate its "source" queue or exchange.
- Policy changes in such a way that a link considers itself no longer necessary.
By increasing reconnect-delay
for upstreams it is possible
to tolerate higher link failure rates. This is primarily relevant
for RabbitMQ installations where a moderate or large number of active links.
Federating Clusters
Clusters can be linked together with federation just as single brokers can. To summarise how clustering and federation interact:
- You can define policies and parameters on any node in the downstream cluster; once defined on one node they will apply on all nodes.
- Exchange federation links will start on any node in the downstream cluster. They will fail over to other nodes if the node they are running on crashes or stops.
- Queue federation links will start on the same node as the downstream queue. If the downstream queue is a replicated one, they will start on the same node as the leader, and will be recreated on the same node as the new leader after any future leader elections.
- To connect to an upstream cluster, you can specify multiple URIs in a single upstream. The federation link process will choose one of these URIs at random each time it attempts to connect.
Securing Federation Connections with TLS
Federation connections (links) can be secured with TLS. Because Federation uses a RabbitMQ client under the hood, it is necessary to both configure source broker to listen for TLS connections and Federation/Erlang client to use TLS.
To configure Federation to use TLS, one needs to
- Use the
amqps
URI scheme instead ofamqp
- Specify CA certificate and client certificate/key pair via URI query parameters when configuring upstream(s)
- Configure Erlang client to use TLS
Just like with "regular" client connections, server's CA should be trusted on the node where federation link(s) runs, and vice versa.
Federation Link Monitoring
Each combination of federated exchange or queue and upstream needs a
link to run. This is the process that retrieves messages from upstream
and republishes them downstream. You can monitor the status of
federation links using rabbitmqctl
and the management
plugin.
Using CLI Tools
Federation link status can be inspected using RabbitMQ CLI tools.
Invoke:
# This command will be available only if federation plugin is enabled
rabbitmqctl federation_status
This will output a list of federation links running on the target node (not cluster-wide). It contains the following keys:
Parameter Name | Description |
type |
|
name | the name of the federated exchange or queue |
vhost | the virtual host containing the federated exchange or queue |
upstream_name | the name of the upstream this link is connected to |
status | status of the link:
|
connection | the name of the connection for this link (from config) |
timestamp | time stamp of the last status update |
Here's an example:
# This command will be available only if federation plugin is enabled
rabbitmqctl federation_status
# => [[{type,<<"exchange">>},
# => {name,<<"my-exchange">>},
# => {vhost,<<"/">>},
# => {connection,<<"upstream-server">>},
# => {upstream_name,<<"my-upstream-x">>},
# => {status,{running,<<"<rabbit@my-server.1.281.0>">>}},
# => {timestamp,{{2020,3,1},{12,3,28}}}]]
# => ...done.
Using the Management UI
Enable the rabbitmq_federation_management
plugin that extends
management UI with a new page that displays federation links in the cluster.
It can be found under Admin
> Federation Status
, or by using the
GET /api/federation-links
HTTP API endpoint.