Wildfly Clustering Singleton MSC Service Sample

This sample installs a simple singleton MSC Service service and simple EJB Singleton Service to monitor the status of the node. And has a simple servlet based health check endpoint to return the status of the node.

1. To run the app inn single standalone instance,

    % mvn wildfly:run
   

   or

2. To test the app in an cluster (change the port as per your need),

   a. Download and extract the wildfly instance to two different locations:

        % wget http://download.jboss.org/wildfly/10.1.0.Final/wildfly-10.1.0.Final.zip
        % unzip wildfly-10.1.0.Final.zip -d wildfly-node1
        % unzip wildfly-10.1.0.Final.zip -d wildfly-node2
   

   b. Start the first wildfly instance:

        % cd wildfly-node1
        % ./bin/standalone.sh -Djava.net.preferIPv4Stack=true --debug 8787 --server-config=standalone-full-ha.xml -Djboss.node.name=node1
   

   c. Start the second wildfly instance:

        % cd wildfly-node2
        % ./bin/standalone.sh -Djava.net.preferIPv4Stack=true --debug 9797 --server-config=standalone-full-ha.xml -Djboss.node.name=node2 -Djboss.socket.binding.port-offset=100
   

   d. Deploy the app in first instance (node1):

        % mvn wildfly:deploy
   

   e. Deploy in app in second instance (node2):

        % mvn wildfly:deploy -Dwildfly.port=10090
   

3. Check the response of the health check end point,

    % curl http://localhost:8080/singleton/health
    % curl http://localhost:8180/singleton/health
   

Testing (Failure Case)

1. Check the health status of the both nodes (both should be consistent):

    % curl http://localhost:8080/singleton/health
    % curl http://localhost:8180/singleton/health
   

2. Let's pause the JVM process

    % jps -l
    {PID list}
   
    % kill -STOP {PID}
   

3. Check the health status of the both nodes (both shouldn't be responding for 60 secs and then it'll be inconsistent):

    % curl http://localhost:8080/singleton/health
    % curl http://localhost:8180/singleton/health
   

4. Resume the paused process after 60 secs and the let both the instances merge its cluster view (wait for few more seconds).

    % kill -CONT {PID}
   

5. Check the health status of the both nodes (both will be inconsistent and both assumes itself as master/active node):

    % curl http/localhost:8080/singleton/health
    % curl http://localhost:8180/singleton/health