I poked around the repository looking for a way to keep an eye on this. I am concerned that the askdata tab is the first one that comes when a user opens a datasource and this might consume a lot of CPU. Has anybody figured out how to monitor this? I am perplexed that Tableau would release this feature and not provide any controls or data on it.
I looked in historical_event_types and there wasn't anything like 'NLP' or 'askdata' but I was able to find it in the http_requests table where the controller starts with '/startAskData/' when the user clicks on the ask data pane and it starts the analysis of the datasource. If you actually ask a question the controller column starts with '/askData/'
Looking further there is a new schema called nlp with tables called datasources, events, fields, and jobs. Looks like these store the analysis results.
To find when datasources have NLP enabled there is a column in the datasources table called nlp_enabled. It defaults to true. I have experimented with writing to the repository and changing this to false because they only way to change it involves a few clicks for each datasource. There is no way I can find to change it for multiple datasources or enable/disable it at the site. The Virtual Server Admin group did have a (undocumented) TSM setting that would disable it for the entire server.
Permissions: if a user has connect and view rights they can use ask data
Looking at the http logs I have about 15-30 daily instances where the analysis is kicked off. 0-2 times a day are users actually using AskData. So it looks like most of the time the analysis is triggered accidentally. How are others handling, controlling, and monitoring askdata?