Jeff Strauss wrote:
... Any automatic way of excluding them?
Or simply just identifying them? I doubt interruptions (e.g. undocking laptop) would be captured by the Tableau Server but maybe the burp would register in a log? Or maybe only if it were in DEBUG mode would such detail possibly emerge?
I agree, it would be nice to know more detail for such outliers. Was it a one-time thing? Does it happen at a certain time/date? Maybe it's a db issue, not a Tableau one? etc.
I will do more investigation within my dev system today to see what may be able to be deciphered. It doesn't appear to be a 1-time thing, no pattern on date/time and it's all extract based.
Suspect that dots on a view are only the "Access View" actions,
not sure about the subsequent bootstraps (not to mention failures).
Besides, cache hits constitute the majority of the lower quartiles,
so the median value could be taken with the grain of salt. imho.
That said, I'm a little bit skeptical about the real value of this view
regarding the performance of the views / dashboards being captured.
I did a bit deeper analysis based on today's data. And I found the following. I am going to talk with one of the users this afternoon and see if we can predictably replicate.
- The data within http_requests is a combination of retrieve and bootstrap request. I believe the view computes render times based on the bootstrap.
- It's an embedded view
- There's multiple requests for the same user, same vizql session
- Things start to go astray when there's overlap of requests
I met with the user that create all the parallel requests. Basically, his business app has a bunch of tabs on it (darn it, I should have gotten a screenprint). Each tab has a different client_id, and there are 8 total. And he switches rapidly between the tabs which each have the dashboard, and the client_id is unique and is passed into the dashboard. Maybe there's some kind of constraint with the dataengine, I really don't know. I just know that there's all kinds of undocumented settings within workgroup.yml and I also know that I'm not reaching cpu stress, ram stress, disk stress (at least according to my performance dashboards).