Last updated at Sun, 05 Nov 2017 16:57:48 GMT

*Mike Roberts is a Logentries customer and Director of Data Analytics at Pluralsight. *

A truly stable and robust Analytics platform is able to support both the analysis of external data *as well as *internal data, or data about its own state. Tableau Software is one such analytics platform and Logentries is the ‘listener’ that makes analyzing Tableau’s system data easier. As a result, our Pluralsight team is better able to understand the finer points of an analytics infrastructure, or what we call DataOps, and release beautiful and functional dashboards and reports for our customers.

There’s enormous value in collecting logs for Tableau Server (and its Desktop product). Much of which would go unnoticed if not for tools like Logentries. For example, we’re able to analyze, in real time, Apache logs to see which views are not loading and, most important, which view (or views) are being downloaded the most via our custom ‘csv’ tag (one of many that we use to parse the logs). This functionality so far has been instrumental in assessing the usefulness of many reports and dashboards. On the other hand, we’re alerted when we receive an http status code that is not 200 or when a query takes longer than a set number of seconds.

The above line chart demonstrates the tags trending over time. The real feedback loop comes from when we leverage the Logentries open API to grab the filtered data, export to csv and then import into Tableau for deeper analysis and dashboarding. One potential long-term benefit of this historical analysis is the ability to be flexible in how we scope our Tableau infrastructure or, simply, to scale linearly with additional processes and resources.

Again, we stress the importance of creating material that is both functional and beautiful; our Pluralsight customers, who enjoy the reports, also want to understand what can be improved with the report’s design. One common request is often how fast a dashboard loads on Tableau Server. And with Logentries ability to parse a JSON payload, and the Open API, we can (with a simple log search function) find out how long each dashboard might take to load and what view in the dashboard might be taking the longest amount of time.

The above bar graph was generated with Logentries and shows what views on Tableau are taking the longest to load. This, again, is invaluable information and is generated live. We’re usually able to reach out to our customers about this before they reach out to us (if they do). The last thing we want our customers to expect is a slow system and, dreadfully, not have any reason to explain it.

For Tableau Server users, understanding the product and its many bits is truly a DataOps-specific task. As data volume continues to grow and become interdependent, operationalizing and analyzing this is paramount to the success of our team (or any company). And the log data, while typically ignored, provides an enormous benefit to us as we continue to analyze the mounds of data. In the end, we’re just connecting the dots from end to end.