There was a known issue with the Tableau Server 9.3.0-.12 versions that was resolved in a later release. Are you running Tableau Server 9.3.0-.12?
The following knowledge base article contains the known issue: VizQLServer Process Doesn't Free Memory | Tableau Software
I hope this helps!
Thanks for the response, I did go through the article. But we are on Tableau server 10.2.2.
We recently upgraded form 10.1.1 to 10.2.2, and we see few of the Dashboards loading very slow and sometimes they give time out errors
Strange thing is that when we restart the server, the dashboard starts loading again quickly.
So my thoughts were if there is some issues with respect to memory management in this particular release.
Hi vineeth v
"Just wanted to know if this is by default that the Vizql server memory will be sustained and the memory will not be released back to system."
No, it will release the memory back to the System. We have seen some issues in version 9.3 but it's not there in 10.2+ atleast.
Has the workbook been republished since the upgrade to Tableau Server 10.2.2? From the new version of Tableau Desktop (10.2.2)? Does this result in the memory being released as expected? I would also suggest opening a case with Tableau Technical Support. - https://www.tableau.com/support/case.
This is actually the VizQL Server Process being restarted. Tableau has a "resource watchdog" (called "SRM") which will purposefully kill a process if it begins using too high of a percentage of the RAM or CPU on Tableau for an extended period of time. What you're seeing is pretty much expected during a load test because by definition you're rendering as many vizzes as you can and therefore "pushing" the VizQL Server Process hard.
If your load test is relatively short (10-20 minutes) and you tend to restart Tableau Server between tests, you typically won't see this. If you run your tests for a longer period and/or don't restart Tableau Server between tests, you'll almost always encounter some of these during a "long haul" test. This is actually Server doing what it's supposed to - making sure that one process doesn't inadvertently consume so many resources that other processes begin to suffer.
You have all the answers Russell Christopher. The SRM link is a real eye-opener. I am hoping to find the exact dashboard that hogged memory during my load test. Next steps for Tableau Server product is to find a way to pipe that info into a report.
If anyone else is looking to monitor server resources, take a look at logshark. I've got a setup guide on medium if you are looking to implement it. One of the output reports is titled ResourceManager and details each time the SRM terminates a process. Logshark makes it very easy to correlate shutdowns with workbooks that were loading. Use the Apache report to see the specific workbooks that were loading at the same time.