Replicate HBase data simultaneously between multiple clusters

Starting from Cloudera 7.2.16.500, 7.2.17.200, and 7.2.18, you can create multiple HBase replication policies between multiple clusters to replicate HBase data. You must consider the limitations before you create a multi-cluster replication scenario. You can use the multi-cluster replication scenario for various use cases.

How multi-cluster HBase replication works

The first-time setup configuration consists of several steps of which one step is to ensure that the source and target cluster use the same credentials.jceks file. Therefore, if multiple supported clusters share the same credentials.jceks file, you can replicate HBase data between them seamlessly using HBase replication policies.

The following image shows a sample multi-cluster HBase data replication scenario and a few possible directions of replication:

It is recommended that you do not replace the credentials.jceks file manually to create a multi-cluster HBase replication scenario. This is because when you create the first HBase replication policy between a pair of clusters, Replication Manager triggers the first-time setup process during which the credentials.jceks file in both the clusters get synchronized as required for HBase data replication.

Limitations

Consider the following limitations before you replicate the HBase data between multiple clusters using HBase replication policies:

An HBase replication policy in a multi-cluster HBase replication setup fails when you use clusters that are part of another independent replication setup. This is because the clusters use a different credentials.jceks file. To use these clusters, you must break the cluster pairing and then create the required HBase replication policies.
tip
To break the cluster pairing, use the POST /dmx/api/clusters/<target cluster crn>/hbase/resetFirstTimeSetup API with the {"sourceCluster": "<source cluster data center>$<source cluster name>"} payload.
Monitor the growing multi-cluster replication network so that it does not get disconnected. This ensures that the credentials.jceks file is the same on all clusters, the replication setup is always consistent, and no existing replication scenarios have to be reset.
The Replication Manager UI does not allow the HBase replication policy creation to proceed if you choose a cluster (as source or target) that is in another first-time setup process. In this instance, you can wait for a few minutes to allow the first-time setup to complete and then create the HBase replication policy.
When you create the first HBase replication policy between two clusters, the first-time setup configuration is initiated. After the configuration completes, the HBase data replication is initiated.
important
In a multi-cluster replication scenario, when you use a source on-premises cluster, you must manually restart the HBase service on the cluster after the first-time setup configuration completes. However, the next time you choose the same source cluster with another target cluster, the manual restart is not required. The same is true for the automatic HBase service restart in Data Hubs where the restart is performed only when the first replication is created with a particular cluster.
The following conditions must be met to use the IDBroker credentials to create multiple HBase replication policies between multiple clusters when the target COD clusters are in separate AWS accounts or when a single AWS Role does not have access to all the required S3 buckets for all HBase target clusters:
- You are using Cloudera Public Cloud 7.2.18.200 or higher versions.
- You choose the Perform Initial Snapshot option, and then specify the custom username in the Export snapshot user field in the Select Source page during the HBase replication policy creation process.

Use cases

Some use cases where you can use the multi-cluster HBase replication scenarios are illustrated below:

Multiple source clusters and a single target cluster. You might have a disaster-recovery use case where you want to use a single COD to back up all the HBase data. The following image illustrates this scenario:
Single source cluster and multiple target clusters. You might have a use case where all the HBase data is located in a cluster and you want to replicate only specific HBase tables to different environments to fulfill specific requirements. For example, QE environments and/or experimentation use case. The following image illustrates this scenario: