Page Comparison

...

1. sys.q.audit.01 - a queue bound to sys.ex.audit
2. sys.q.dead.letter.01 - a queue bound to sys.ex.auditdead.letter
3. sys.q.unrouted.01 - a queue bound to sys.ex.auditunrouted
4. q.type1 - a queue that gets 'type1' messages from ex.router.main
5. q.type2 - a queue that gets 'type2' messages from ex.router.main

...

1. unroutedMessagePolicy - applies to all Exchanges starting with 'ex.router'. If the message is unouted unrouted by the Exchange, it will be sent to the 'alternate exchange' : sys.ex.unrouted
2. messageTimeoutPolicy - applies to all Queues starting with 'q'. If the message is left unprocessed on the Queue after 30 seconds, it will be sent to the 'dead-lettter-exchange' : sys.ex.dead.letter. 

Notes about the Demo configuration

The naming of resources is important - this ensures application Queues and Exchanges are separated out from System Queues and Exchanges. For example messages on the sys

...

.q.dead.letter.01 don't have a TTL but those on q.type1 and q.type2 do.

We've had to use 2 exchanges ex.entry.point and ex.router.main. The reason for this is that if have a single exchange from which we copy every message to audit - this will stop messages from appearing unrouted even if they only go to the audit queue.

The functionality covered by the 2 policies could be defined on each Queue/Exchange but this is easy to get wrong. Using policies reduces overhead and increases quality.

Instructions for setting up the demo and exploring its features

...

- other exchanges start with "ex". This naming pattern is important because of the Policies which apply extra functionality to groups of Queues/Exchanges.

$ rabbitmqadmin --user=$RBT_USER --password=$RBT_PASSWD --vhost=$VHOST list exchanges | grep -v amq | grep -v direct
+--------------------+---------+
|        name        |  type   |
+--------------------+---------+
| ex.entry.point     | fanout  |
| ex.router.main     | topic   |
| sys.ex.audit       | fanout  |
| sys.ex.dead.letter | fanout  |
| sys.ex.unrouted    | fanout  |
+--------------------+---------+

...

The first 2 are fanout bindings from ex.entry.point to sys.ex.audit and  and ex.router.main

The next 2 are topic bindings from ex.router.main to q.type1 for routing key 'type1' and q.type2 for routing key 'type2'. Notice - no binding for 'type3'

The last 3 are fanout bindings for each of the 'sys' exchanges to their message queues.

davidhay@localhost RABBITMQADMIN-DEMO (master) $ rabbitmqadmin --user=$RBT_USER --password=$RBT_PASSWD --vhost=$VHOST list bindings | grep -v '^|  '
+--------------------+----------------------+----------------------+
+--------------------+----------------------+----------------------+
| ex.entry.point     | ex.router.main       |                      |
| ex.entry.point     | sys.ex.audit         |                      |
| ex.router.main     | q.type1              | type1                |
| ex.router.main     | q.type2              | type2                |
| sys.ex.audit       | sys.q.audit.01       |                      |
| sys.ex.dead.letter | sys.q.dead.letter.01 |                      |
| sys.ex.unrouted    | sys.q.unrouted.01    |                      |
+--------------------+----------------------+----------------------+

...

In addition to Exchanges, Queues and Bindings - this demo relies on 2 Routing "Policies". These help provide extra routing for unprocessed and unrouted messages to groups of Exchanges/Queues.

The policy 'unroutedMessagePolicy' applies to exchanges starting with 'ex.router' - any unrouted message gets sent to 'sys.ex.unrouted'

The policy 'messageTimeoutPolicy' applies to queues starting with 'q' - any message still on a queue after 30seconds 30 seconds is sent to 'sys.ex.dead.letter'

Note: because of these policies - it's important to name the queues and exchanges carefully.

This shows the 2 policies described abovein the configuration.

rabbitmqadmin --user=$RBT_USER --password=$RBT_PASSWD --vhost=$VHOST list policies | grep "^| $VHOST"
| vhDemo01 | unroutedMessagePolicy | exchanges | {"alternate-exchange": "sys.ex.unrouted"}                            | ^ex.router.* | 0        |
| vhDemo01 | messageTimeoutPolicy  | queues    | {"message-ttl": 30000, "dead-letter-exchange": "sys.ex.dead.letter"} | ^q.*         | 1        |

...

It should be on q.type1 and a copy of on sys.q.audit.01. This demonstrates routing based on message type and auditing.

...

If we wait 30 seconds and check again. The message should timeout on q.type1 and be sent to sys.q.dead.letter.01. This demonstrates the messageTimeoutPolicy.

rabbitmqadmin --user=$RBT_USER --password=$RBT_PASSWD --vhost=$VHOST list queues
+----------------------+----------+
|         name         | messages |
+----------------------+----------+
| q.type1              | 0        |
| q.type2              | 0        |
| sys.q.audit.01       | 1        |
| sys.q.dead.letter.01 | 1        |
| sys.q.unrouted.01    | 0        |
+----------------------+----------+

...

It should be on q.type2 and a copy of on sys.q.audit.01. This demonstrates routing based on message type and auditing.

...

If we wait 30 seconds and check again. The message should timeout on q.type2 and be sent to sys.q.dead.letter.01. This demonstrates the messageTimeoutPolicy.

rabbitmqadmin --user=$RBT_USER --password=$RBT_PASSWD --vhost=$VHOST list queues
+----------------------+----------+
|         name         | messages |
+----------------------+----------+
| q.type1              | 0        |
| q.type2              | 0        |
| sys.q.audit.01       | 2        |
| sys.q.dead.letter.01 | 2        |
| sys.q.unrouted.01    | 0        |
+----------------------+----------+

...

The message should end up on sys.q.unrouted.01 because  because when it arrived at ex.router.main there is no mapping for 'type3' messages so it gets sent to sys.ex.unrouted which forwards to sys.q.unrouted.01. This demonstrates the unroutedMessagePolicy.

...