1

We are currently using ActivePivot 4.3 on a horizontal distribution.

We split our data per historical days. Even if the loading is fast, we do not want to let users have access to a partially loaded node.

We would like to have a node outside the cluster while it's loading, and join it when it's ready. Also we might want to keep it out of the cluster if it is under maintenance. In order to achieve that, we need some control on how a node can join/leave the cluster.

I believe there is some control trough JMX, however, we would like to have these controls accessible programatically, ideally through a webservice.

How can we implement that ?

Steve Czetty
  • 6,147
  • 9
  • 39
  • 48
bendoubs
  • 11
  • 2

1 Answers1

1

With ActivePivot 4.3.3, you can achieve it like this:

To prevent a node from joining the cluster before it's fully loaded, you can set the "autoStart" distribution property to false.

In your schema.xml file, add the following code

...
    <distributionDescription>
        <distributedPivotId> xxxx </distributedPivotId>
        <clusterId> xxxx </clusterId>
        <distributionType> xxxx </distributionType>
        <properties>
            <entry key="autoStart" value="false" />
        </properties>
    </distributionDescription>
...

Then, the messenger component, responsible for communication through the network, won't be started. You have to do it manually.

To start this component, you have to call its "start()" method. If you don't want to use JMX tools and do it programatically, you have to use the "ActivePivotManager" component. It provides methods to get the several instances of ActivePivot: use it to get the desired distributed ActivePivot.

Finally, use the "getMessenger()" method from the distributed ActivePivot to get the messenger component, and start it. The considered node will then join the cluster.

Assuming you have the manager, your code should look like the following :

ADistributedActivePivot myDistributedPivot = (ADistributedActivePivot) manager.getActivePivots().get("myDistributedPivot"); // Change it to the Id of your distributed pivot
myDistributedPivot.getMessenger().start();

To make a node join/leave the cluster, you can freely use the "pause()" and "resume()" method from its messenger.

fsa
  • 36
  • 2