Tableau Server Administration & Governance

Leading up to my Tableau Conference presentation, I wanted to put together a series of blog posts that discuss the three pillars of my discussion and approach to self-service analytics. The first pillar is Training and can be found here. Without a trained set of super users, the Security structure and Governance model you will develop doesn't matter because no one is using it. With a trained set of Tableau developers, a simple, scalable Security model is foundational to the efficient use of the technology and is discussed in this post. The third and final pillar is the monitoring and governance of the action happening on Tableau Server. I like to utilize the Postgres database that lives behind Tableau Server to check the rules outlined in my previous post in addition to providing proactive support of our users.

Below you will find a link where you can download a pre-built Tableau workbook from Dropbox. The workbook was built with live connections to a Tableau Server instance. When opening the workbook, you will be prompted for credentials. In the upper right corner you can edit the connection and point it at your Tableau Server instance. I built this in Tableau 10.5 and it may not operate as expected in past or future versions. I provide these without warranty or the promise of support.

These have been generated over the course of many years. One of the dashboards was adapted from Mark Jackson's blog called Tableau Zen whose post helped me start down this path. You may need to modify the views for your own specific needs. When doing so I recommend looking at Tableau's data dictionary. I hope to add to this workbook in the future as there are additional views I have created that would be of use. Please leave a note below about what you would like to see of what would be valuable to you!

The first dashboard provides some high level information about the sheer number of objects on your server as well as a few quarter-to-date statistics. Content creators are essentially individuals with Tableau Desktop that have published a workbook to the server. It's an attempt at determining how many desktop licenses there are.

The second dashboard shows the permissions that have been setup on the server. I have built this assuming the criteria and methodology I laid out in this post. The top three sheets check whether certain rules have been met: permissions must be locked to a project, permissions must be granted to a group (not a user), and all users must be added to the server with a minimum site role of publisher. If these rules are broken, the objects that need to be updated will be listed. Below are the permissions set for each project.

The third dashboard is my take on Mark Jackson's and checks for subscription or extract failures. I subscribe to this dashboard and receive it every evening. Below are the details of each job. By selecting the sliver of a failure above, the failed jobs are shown below. This is a great way to provide details without being an admin.

The fourth dashboard shows when refreshes occur (and the frequency) over a 4-week period. There is a field in this table called 'Started At' that is set to the GMT time zone. I did a basic calculation to subtract 5 hours to essentially transform it to the central time zone. Down below list the refreshes that take the longest as well as the largest datasources and workbooks. The bars are colored to indicate the objects above the 95th percentile.

The fifth dashboard looks pretty boring, but is pretty cool. It lists any workbook or datasource objects that have not been used in the last 6 months. You can select a bar and in the tooltip see a 'Send Email' menu action. Clicking this opens an Outlook email with pre-populated content such as the owner's email (in the To spot), a subject (what object type and it's ID), and the body. This is probably satisfactory on its own, but we actually use it to also BCC the email to our team's Jira board. The process creates a task in our backlog to remove the object in question. Being super nerdy we bring that Jira data back into this workbook and color code the bar to indicate a Jira ticket has been created and it is or is not within the time frame for deletion.

Enjoy! Good luck! Report back with ideas, comments, or issues.