Installing Tableau Server

Tableau Server was installed directly in an AWS private subnet for internal use. There were three installation methods available—direct install, AWS Quick Start, and AWS Marketplace—but since Quick Start places a single node in a public subnet and the Marketplace AMI was not up to date, direct installation was chosen.

Before selecting an instance, the recommended specs for AWS were checked. T2 instance types were excluded, and the recommendation of 8 CPU cores (16 AWS vCPUs), 64GB RAM was referenced; since the number of users was not large, m5.2xlarge was chosen as a starting point. It was also emphasized that dynamically expanding EBS capacity after installation can cause problems, so it's important to allocate sufficient disk capacity from the start.

Domain and security were also prepared together. The domain was set up in Route53, and SSL was configured to secure communication between Tableau Desktop and the server. Tableau Server uses ports 80 and 443 by default, and the TSM admin page uses port 8850.

The installation was performed on Ubuntu 18.04. After downloading and installing the installer, TSM (Tableau Services Manager) was initialized, and following the web-based setup guide allowed access to the server. Account creation was required on first access, and once the setup was completed successfully, Tableau Server could be accessed from Chrome.

The server is composed of multiple processes.

• Gateway: Web interface, uses Apache internally
• VizPortal: Handles web requests and REST API
• Interactive / Non-Interactive Microservice Container: Manages internal microservices
• VizQL Server: Core of query execution and view rendering
• Cache Server: Query cache, uses Redis internally
• Cluster Controller: Monitors and restarts key functions
• Search and Browse: Search and filter functionality, uses Solr internally
• Backgrounder: Handles background tasks such as extract updates
• Data Server / Data Engine: Data delivery and the Hyper-based engine
• File Store: Stores extract files
• Repository: Stores users, permissions, projects, and metadata, uses PostgreSQL internally
• Tableau Prep Conductor: Runs Prep flows uploaded to the server, requires a Data Management license
• Ask Data / Elastic Server: The "Ask Data" feature and search index, uses Elasticsearch internally
• Messaging Service: Communication between microservices, uses ActiveMQ internally
• TSM Controller: REST API for controlling Tableau Server
• License Server: License management

As this many processes run on a single node, high specs are required, and aside from the core features VizQL and Hyper, a number of open-source components are actively used. The web-based services are mostly written in Java.

Finally, connecting to a specific database requires installing the corresponding driver. Since PostgreSQL is used internally, the related driver was installed, and the installation was then completed in a single-node configuration. However, if HA or load balancing is needed, additional configuration is required.