Master/Slave - Shared Store

A-MQ 7 has the ability to run in a master/slave (ie, active/passive) configuration so that if an active node goes down, another can immediately take its place and continue serving clients and processing messages. There are a few different options that can be used to achieve this:

  • Shared store
  • Replicated store (ie, "shared nothing")
  • Co-located

For the purpose of this lab, we will focus on the "Shared store" option for persistence, and use the "Static" option for clustering.

Prerequisites

Download and install the A-MQ 7 broker package as described in Lab 1.

Creating the brokers

$ $AMQ_HOME/bin/artemis create brokers/master --name master --user admin --password admin --allow-anonymous
$ $AMQ_HOME/bin/artemis create brokers/slave --name slave --user admin --password admin --allow-anonymous --port-offset 1

Note: We could have passed in some extra arguments to configure our master/slave pair with some defaults, but instead we're going to manually configure them via the broker.xml file so we can become familiar with the available options.

Editing the configuration

Master

  1. Open up the brokers/master/etc/broker.xml file in your favorite text editor.

  2. Modify the existing paging-directory, bindings-directory, journal-directory, and large-messages-directory elements to point to a shared mount. In the real world, this would likely point to a directory on a SAN that is mounted on both the master and slave machines.

<paging-directory>../sharedstore/data/paging</paging-directory>
<bindings-directory>../sharedstore/data/bindings</bindings-directory>
<journal-directory>../sharedstore/data/journal</journal-directory>
<large-messages-directory>../sharedstore/data/large-messages</large-messages-directory>
  1. Add the following elements anywhere under the <core> element:
<connectors>
  <connector name="master-connector">tcp://localhost:61616</connector>
  <connector name="slave-connector">tcp://localhost:61617</connector>
</connectors>

<cluster-user>admin</cluster-user>
<cluster-password>admin</cluster-password>

<cluster-connections>
  <cluster-connection name="static-cluster">
    <connector-ref>master-connector</connector-ref>
    <static-connectors>
      <connector-ref>slave-connector</connector-ref>
    </static-connectors>
  </cluster-connection>
</cluster-connections>

<ha-policy>
  <shared-store>
    <master>
      <failover-on-shutdown>true</failover-on-shutdown>
    </master>
  </shared-store>
</ha-policy>
  1. Start the broker:
$ ./brokers/node1/bin/artemis run

Slave

  1. Open up the brokers/slave/etc/broker.xml file in your favorite text editor.

  2. Modify the existing paging-directory, bindings-directory, journal-directory, and large-messages-directory elements to point to a shared mount. In the real world, this would likely point to a directory on a SAN that is mounted on both the master and slave machines.

<paging-directory>../sharedstore/data/paging</paging-directory>
<bindings-directory>../sharedstore/data/bindings</bindings-directory>
<journal-directory>../sharedstore/data/journal</journal-directory>
<large-messages-directory>../sharedstore/data/large-messages</large-messages-directory>
  1. Add the following elements anywhere under the <core> element:
<connectors>
  <connector name="master-connector">tcp://localhost:61616</connector>
  <connector name="slave-connector">tcp://localhost:61617</connector>
</connectors>

<cluster-user>admin</cluster-user>
<cluster-password>admin</cluster-password>

<cluster-connections>
  <cluster-connection name="static-cluster">
    <connector-ref>slave-connector</connector-ref>
    <static-connectors>
      <connector-ref>master-connector</connector-ref>
    </static-connectors>
  </cluster-connection>
</cluster-connections>

<ha-policy>
  <shared-store>
    <slave>
      <failover-on-shutdown>true</failover-on-shutdown>
      <allow-failback>true</allow-failback>
    </slave>
  </shared-store>
</ha-policy>
  1. Start the broker:
$ ./brokers/node2/bin/artemis run

Testing

  1. Open up a new terminal window and run the following command:
$ $AMQ_HOME/bin/artemis producer --verbose --user admin --password admin --message-count 100 --url 'tcp://localhost:61616'

  1. Shut down the master broker (ie, by hitting ctrl-c on its terminal). You should see the slave broker log a few messages indicating that it has come online (this might take a couple of seconds).

  2. In the same terminal window that you ran the producer command, run the following command:

$ $AMQ_HOME/bin/artemis consumer --verbose --user admin --password admin --message-count 100 --url 'tcp://localhost:61617'

All of the messages that you produced to the master broker in step 1 should be consumed from the slave broker.

  1. (optional) Restart the master broker and you should see that it will take over and the slave will shut down/go back into slave mode (this might take a couple of seconds).