A highly available installation of Tableau Server is a distributed installation that is designed to maximize the availability of Tableau Server.
If you’re configuring a Tableau Server installation for high availability (HA), the steps you perform are all designed to build in redundancy, thus reducing your potential downtime. The four areas of Tableau Server that require redundancy are the data engine, repository (pgsql), and gateway processes, along with the primary node, which runs the server's licensing component. Because there must always be one active of the repository process, configuring the cluster is a multi-phased procedure that requires the primary Tableau Server to be stopped and restarted at certain points so that settings can take effect. For exact steps, see Configure for Failover and Multiple Gateways and Use a Backup Primary. See Distributed Requirements as well.
The minimum supported configuration for high availability is a three-node system. This includes a primary server node to run licensing, and two worker nodes to host the main processes. You can increase reliability of the system by adding a fourth computer to serve as a backup primary. If you run a gateway process on all nodes, it also makes sense to use a load balancer for the gateways.
A Single Server System
After you install the primary Tableau Server, it is running at least one instance of all server processes. This is the most basic configuration of Tableau Server. It has no redundancy.
Here’s what the Process Status table on the Server Status page typically looks like for a single-server system (not all processes shown above appear on the Status page):
To build in redundancy, you need to add additional servers to host copies of the repository and data engine/file store processes. In addition, to reduce the system’s vulnerability, you can run multiple gateways, and the primary should be isolated on its own node, ideally running as few of the server processes as possible. The fewest number of computers required to achieve this is three (see A Three-Node System).
A three-node system helps you reduce the primary's vulnerability:
This configuration would look like the following Process Status table on the Server Status page.
In a three-node cluster, the Data Engine and Repository processes have been moved from the primary to a worker, and the primary is only running the Gateway and Search & Browse processes. In this configuration, if your active worker fails, the passive worker automatically becomes active. Exactly how to create this three-node cluster, including how to add the workers and remove the processes from the primary, is described in Configure for Failover and Multiple Gateways. (Licensing functionality is integral to the primary and cannot be removed, so it is not displayed on the Status page. Cluster Controller and Coordination Service are installed on all nodes as part of the "base install" and are not configurable. Coordination Service does not show on the Status page and Cluster Controller only displays if there are two or more nodes in the cluster.)
There are still two things you can do to improve this three-node cluster: 1) add a load balancer to interface with the three active gateways, and 2) create a backup to address the single point of failure: the primary. See the topics below for details.
Add a Load Balancer
At this point, all three nodes have gateways, which are used to route requests to available server processes. Unlike the repository process, there aren't active and standby gateways. All gateways are active. To further reduce your cluster's potential for downtime, you should configure a load balancer.
Add a Backup Primary
Adding a backup primary provides a safeguard for your system. The backup primary is an additional server added to the system to be ready if your primary fails. While it is not an active server, after you complete the first set of steps in Use a Backup Primary, it is ready to be activated. While the backup primary needs to be licensed during installation, it does not count as one of the three environments allowable under the Tableau EULA.
Here’s what the system looks like with a backup primary:
The Process Status table for the configuration shown above looks the same as for a three-node system. If the primary fails and you perform the steps for the backup primary to take over, your system is back online using the new primary:
Tableau Server service license check
A number of processes are installed when you install Tableau Server. Some of these processes are dependent on the existence of a valid Tableau Server license while other installed processes do not. The subset of Tableau Server that require a valid Tableau Server license are considered "licensed processes."
When a licensed process starts or restarts, the process checks with the Tableau Server License Manager service on the
ProgramData\Tableau\Tableau Server\data\tabsvc\logs\licensing folder