Premium GEN2 | Overloading & Diagnosing Problems

[Anonymous]

Hello,

I've recently upgraded to Premium GEN2 to benefit from the Autoscaling option available. I have done this by opting into the 'Preview' - assume that general availablity hasn't been released yet.

Despite autoscaling up by a couple of extra cores, we're still hitting the limit and performance is taking a hit. I've been using the new App to highlight the artefacts causing the problem and one suspicion I have is that end users are using some of the reports to select huge amounts of data and export this. If this is the case, is there any way to identify to users submitting resource intensive queries to the dataset? I don't want to disable export as it will cause more grief than it's worth with stakeholders.

Also noticed that the options to configure things like Query timeout that used to be in the workload settings are no longer available. My theory was that we could reduce the timeout limit and block expensive queries that way, but doesn't appear to be an option anymore.

Any advice on how to go about diagnosing and shutting down the source of the overloading would be appreciated - I can't believe that a few hundred users would justify upgrading to the next level of capacity...

[GilbertQ · Accepted solution]

Hi @Anonymous 

This can certainly happen if people are looking to export large amounts of data because this is not how PBI is meant to be used. Yes it can be done but it should not be always done.

The good news is that you can now enable log analytics and you can get all the details of long running queries (which means big exports) which you can find out which person is running the query and what query they are running.

Using Azure Log Analytics in Power BI (Preview) - Power BI | Microsoft Docs

[Anonymous]

Thanks @GilbertQ - I'll investigate whether we can enable this for some of the likely culprits. Any ideas on what to be on the lookout for in the log analytics data? We've used the M365 audit log data previously to identify the number of - for example - export operations being performed, but that doesn't differentiate between a small export of a few rows or exporting an entire dataset. 

Presumably log analytics gives a little more detail?

[GilbertQ]

Yeah the log analytics will tell you how long the query took to run, which could give an indication of larger exports. It will also give you the DAX query, so you will be able to see what they are doing too!