As a test, it might be worth republishing a few of the workbooks and datasources using Tableau Desktop 10.3. There have been a few recent threads but when you upgrade Tableau Server, it doesn't actually upgrade the workbooks themselves. They'll remain in the last version they were updated in either via a publish/republish event or the owner of the workbook using WebEdit and then saving the workbook.
Dashboards that we did performance bench mark between 10.1.3 vs 10.3 were republished with Tableau Desktop 10.3.
A little advice for next time... Always burn test new versions of Tableau
Server in dev for at least 1-2 weeks before upgrading your production
server. Testing would have identified this issue before it was rolled out
and it wouldn't be causing a production issue.
On Jun 12, 2017 5:46 PM, "Kalyan Renganathan" <firstname.lastname@example.org>
We have not upgraded 10.3 to Production. It is SandBox (10.3) Vs Production (10.1.3).
I have the same issue. I ran some Tabjolt test scenarios on my QA server (same configuration as production) and after I upgraded QA from 10.0 to 10.3, when repeating the same tests, I have seen a very pronounced degradation in performance. For the same 30 minute test, Average Tests per Second dropped from 1.17 to .29 and Average Response Time went from 17 seconds to 73 seconds. I'm seeing similar behavior when executing individual dashboards interactively on the server.
My Question is how the Tableau public performs well with Same 10.3? Is there any configuration changes required to tune the performance? I tested with sub prod environment with HA (1 Primary , Worker Node1 Active Repository with 2 VizQL, 2 DE, 2 DS and other processes except backgrounder, 1 Passive Repository similar to Worker Node 1 and Worker3 dedicated for Bacgrounders.).
Please share your learning in case if you come across any.