The  can be installed across multiple nodes for high availability failover. This section describes the general process for installing the platform across multiple, highly available nodes.

NOTE: This section applies to deployments of the within the enterprise infrastructure.

The  can also integrate with a highly available Hadoop cluster. For more information, see Enable Integration with Cluster High Availability in the Configuration Guide. 

Limitations

The following limitations apply to this feature:


For more information, see Configure for High Availability in the Configuration Guide.

Overview

The  supports an Active-Active HA deployment model, which works well at scale. The architecture features a single load balancer sitting in front of multiple nodes running the . Each node:

 

Job interruption

In case of a failover event, any in-progress job should be marked as failed. 

Failover events/scenarios around jobs: 

#JobEventResulting job state
1In progressThe batch job runner is fine, but executor running the job fails.

Failed

2In progressThe batch job runner or the node dies.

In Progress

3QueuedThe batch job runner or the node dies.

In Progress1 

4PendingThe batch job runner or the node dies.

In Progress1 2 

1 It may not be "In Progress". However, the job has not truly failed.

2 A nuance around #3. There is a feature flag that can be enabled and is enabled by default, which causes pending jobs to be marked as failed on (re)start of batch job runner. However, because this feature indiscriminately marks all pending jobs as failed, it cannot be safely enabled in an environment that has multiple running batch job runners. 

Installation Topography

The  supports a single load balancer placed in front of multiple nodes, each of which runs the same version of . Content between nodes is shared using an NFS resource mount.

Shared resources:

Each node shares the following resources:

Order of Installation

Steps:

  1. All nodes must meet the system requirements. See System Requirements in the Planning Guide.

  2. All nodes must have the appropriate ports opened. See System Ports in the Planning Guide.

  3. Install the databases.

    NOTE: The databases must be installed in a location that is accessible to all nodes.

    NOTE: When installing databases for high availability access, you should deploy standard access and replication techniques that are consistent with the policies of your enterprise.

     

    See Install Databases in the Databases Guide.

  4. Complete the installation process for the server node.

    NOTE: After install, do not start the .

    See Install Software.

  5. Repeat the above process for each of the client nodes.

  6. The software is installed on all nodes. No node is running the software.

Configuration

Additional configuration is required.

NOTE: Starting and stopping the platform in high availability mode requires additional steps.

For more information, see Configure for High Availability in the Configuration Guide.