3.8. Replicating from a Cluster to a Datawarehouse

You can replicate data from an existing cluster to a datawarehouse such as Hadoop or Vertica. A replication slave node handles the datawarehouse loading by obtaining THL from the cluster. The configuration of the cluster needs to be changed to be compatible with the required target applier format.

The cluster-slave deployment works by configuring the cluster replication service in heterogeneous mode, and then replicating out to the slave that writes into the datawarehouse by using a cluster alias. This ensures that changes to the cluster topology (i.e. master switches during a failover or maintenance) still allow replication to continue effectively to your chosen datawarehouse.

The datawarehouse may be installed and running on the same host as the replicator, "Onboard", or on a different host entirely, "Offboard".

Figure 3.6. Topologies: Replication from a Cluster to an Offboard Datawarehouse

Topologies: Replication from Cluster to Datawarehouse

Below is a summary of the steps needed to configure the cluster-slave topology, with links to the actual procedures included:

  1. Install or update a cluster, configured to operate in heterogeneous mode.

    In our example, the cluster configuration file /etc/tungsten/tungsten.ini would contain two stanzas:

    • [defaults] - contains configuration values used by all services.

    • [alpha] - contains cluster configuration parameters, and will use topology=clustered to indicate to the tpm command that nodes listed in this stanza are to be acted upon during installation and update operations.

    For more details about installing the source cluster, please see Section 3.8.2, “Replicating from a Cluster to a Datawarehouse - Configuring the Cluster Nodes”.

  2. Potentially seed the initial data. For more information about various ways to provision the initial data into the target warehouse, please see Section 3.9, “Migrating and Seeding Data”.

  3. Install the slave replicator:

    In our example, the slave configuration file /etc/tungsten/tungsten.ini would contain three stanzas:

    • [defaults] - contains configuration values used by all services.

    • [alpha] - contains the list of cluster nodes for use by the slave applier service as a source list. This stanza will use topology=cluster-alias to ensure that no installation or update action will ever be taken on the listed nodes by the tpm command.

    • [omega] - defines a replicator slave applier service that uses topology=cluster-slave. This service will extract THL from the cluster nodes defined in the relay source cluster-alias definition [alpha] and write the events into your chosen datawarehouse.

    For more details about installing the slave replicator, please see Section 3.8.3, “Replicating from a Cluster to a Datawarehouse - Configuring the Cluster-Slave”.