Configuring Technology Specific Details

Technology Configuration

The technology specific configuration parameters can be easily accessed and updated through Configure and Map Nodes option. Also, there are two different options available for deployment namely Deploy and Register. It should be noted that node mapping is not possible when registering a technological component.

Default Values for Configuration

The configuration parameter inputs are pre-populated with default values that are recommended as per best practices. The default values also helps to minimize the number of inputs for quickly deploying a cluster.

Deploying or Registering a Component

You can either Deploy the required components (technologies) afresh or use the already existing components under Register mode.

Prerequisites for Registering a Component

Following is the scope and definition for registering a cluster. It can be counted as prerequisites.

Common:

1. Cluster must be a valid and fully functional cluster

2. For components (Zookeeper and Hadoop), user must enable JMX via a port to enable Ankush monitoring before registration

3. To enable Ganglia monitoring with component, if any technology specific process requires a restart, then user have to perform it manually

4. PATH and JAVA_HOME environment variables must be declared on all the nodes

5. Registration of components deployed are supported only through tar.gz/zip packages

6. Component configuration and Logs folder must be at default locations inside the extracted package

7. If Ganglia is already installed and we choose to register Ganglia, Ankush assumes that it's not integrated with the component. During registration, Ankush will configure Ganglia for all those components

8. Cluster deletion after successful registration – It will stop all the processes and will delete all the component specific directories from the nodes

9. Cluster deletion after registration failure - It will restore the node’s state to the state before registration

10. Unregister is not supported for successful registered clusters

Hadoop:

1. JMX data should be available on NameNode http Port (http://:/jmx)

2. For Hadoop1, NameNode and JobTracker should be on same node

3. For Hadoop2, HA is supported only with QJM method (with journal node process management)

4. For Hadoop2, Job History Server and WebAppProxyServer should use default http as well as RPC ports

Ganglia:

1. Ganglia services should be running on all nodes

2. Validating gmetad conf, gmond conf and RRD directory path existence

3. Graph will show all the metrics that ganglia is monitoring

Note: In many of the path related configuration parameters, the default value that appears contains “$user”. It should be noted that, it automatically gets updated to “username” based on the value provided in user name input of node authentication.