Configuration

When creating an exchange the name of an AE can be optionally supplied in the exchange.declare method's arguments table by specifying a key of 'alternate-exchange' and a value of type 'S' (string) containing the name.

When an AE has been specified, in addition to the usual configure permissions on the declared exchange, the user needs to have read permissions on that exchange and write permissions on the AE.

For example:

Map<String, Object> args = new HashMap<String, Object>();
args.put("alternate-exchange", "my-ae");
channel.exchangeDeclare("my-direct", "direct", false, false, false, args);
channel.exchangeDeclare("my-ae", "fanout");
channel.queueDeclare("routed");
channel.queueBind("routed", "my-direct", "key1");
channel.queueDeclare("unrouted");
channel.queueBind("unrouted", "my-ae", "");

In the above fragment of Java code we create a direct exchange 'my-direct' that is configured with an AE called 'my-ae'. The latter is declared as a fanout exchange. We bind one queue 'routed' to 'my-direct' with a binding key of 'key1', and a queue 'unrouted' to 'my-ae'.

Operation

Whenever an exchange with a configured AE cannot route a message to any queue (or the message was marked as immediate and it cannot be delivered to any consumer) then it publishes the message to the specified AE instead. If that AE does not exist then a warning is logged.

For example:

channel.basicPublish("my-direct", "key2", null, "test".getBytes());
GetResponse r2 = channel.basicGet("unrouted", true); //returns message
channel.basicPublish("my-direct", "key1", null, "test".getBytes());
GetResponse r1 = channel.basicGet("routed", true); //returns message

Here we first publish a message to 'my-direct' with a routing key of 'key1'. That message is routed to the 'routed' queue, in accordance with the standard AMQP behaviour. However, when publishing a message to 'my-direct' with a routing key of 'key2', rather than being discarded the message is routed via our configured AE to the 'unrouted' queue.

When an AE cannot route a message, it in turn publishes the message to its AE, if it has one configured. This process continues until either the message is successfully routed, the end of the chain of AEs is reached, or an AE is encountered which has already attempted to route the message.

The semantics of the 'mandatory' and 'immediate' flags carries through to AEs. So, for example, if a sender specified the 'mandatory' flag on publication of a message, and that message could not be routed by the original exchange, but was then routed to a queue via an AE, then no basic.return is issued.

Message properties are carried through when messages are routed via AEs. For example, when a message that was marked as persistent is routed to some queue via an AE then it does get persisted.

The queues to which a messages is routed from an AE participate in AMQP's tx transactions, i.e. if an unroutable message was published inside a tx then it will only appear in the queues when that tx is committed.

Usage

The memory threshold at which the flow control is triggered can be adjusted by editing the rabbitmq.config file (in the appropriate location for your platform, as discussed in the installation guide). The example below sets the threshold to the default value of 0.4:

[{rabbit, [{vm_memory_high_watermark, 0.4}]}].

The default value of 0.4 stands for 40% of installed RAM or 40% of available virtual address space, whichever is smaller. E.g. on a 32-bit Windows platform, if you have 4GB of RAM installed, 40% of 4GB is 1.6GB, but 32-bit Windows normally limits processes to 2GB, so the threshold is actually to 40% of 2GB (which is 820MB).

A value of 0 disables the memory monitor and throttling of producers via channel.flow.

The memory limit is appended to the RABBITMQ_NODENAME.log file when the RabbitMQ server starts:

=INFO REPORT==== 29-Oct-2009::15:43:27 ===
Memory limit set to 2048MB.

Versions of Erlang prior to R13B had a bug that meant they do not reliably detect the amount of memory installed on Windows platforms. If RabbitMQ is running under Windows on a version of Erlang prior to R13B, then it does not attempt to automatically set a threshold. Instead, it behaves as if it's unable to recognise the platform: see below.

Unrecognised platforms

If the RabbitMQ server is unable to recognise your system, or if you're using Windows and a version of Erlang prior to R13B, it will append a warning to the RABBITMQ_NODENAME.log file. It then assumes than 1GB of RAM is installed:

=WARNING REPORT==== 29-Oct-2009::17:23:44 ===
Unknown total memory size for your OS {unix,magic_homebrew_os}. Assuming memory size is 1024MB.

In this case, the vm_memory_high_watermark configuration value is used to scale the assumed 1GB RAM. With the default value of vm_memory_high_watermark set to 0.4, RabbitMQ's memory threshold is set to 410MB, thus it will throttle producers whenever RabbitMQ is using more than 410MB memory. Thus when RabbitMQ can't recognize your platform, if you actually have 8GB RAM installed and you want to invoke channel.flow to throttle producers when the RabbitMQ server is using above 3GB, set vm_memory_high_watermark to 3.

It is advised you do not set the threshold above 50% of your installed RAM, and to note that under Windows, it is commonly the case that a maximum of 2GB memory is available to RabbitMQ.