Storm High Availability and Failover - Real-time Analytics with Storm and Cassandra - page 107

Building high availability of components
Now we are at an opportune juncture to look for high availability of various components in
the cluster. We will do this as a series of exercises wherein we assume that each component
is installed in the clustered mode and more than one instance of it exists in the ecosystem.
The high availability of RabbitMQ can be checked only after you have a mirrored queue in
place. Let's assume:
• We have two nodes in the RabbitMQ cluster: node1 and node2
• MyExchange is the name of the exchange that is created for the purpose of this
exercise
• MyQueue is a mirrored queue that is created for this exercise
Next, we will just run the fixedEmitter code we created in the Creating a RabbitMQ
feeder component section. Now perform the Litmus test:
• Let's assume the queue MyQueue has 100 messages
• Now bring down node2 (this means, one node on the cluster is down)
• All the 100 messages will be retained and will be visible on the console; node1
fills in when there is an absence of node2
This behavior ensures that services are not disrupted even if a node in the cluster goes
down.