The operation was throttled by Power BI because the server is under memory pressure in Pro

[SofBL]

Recently a user in my organization started to show me messages of reports that reside in a Pro workspace capacity. 

These reports failed to show their visuals stating: 
Error fetching data for this visual.

The operation was throttled by Power BI because the server is under memory pressure. Please try again later. 

The technical details reveal: 
Underlying Error QueryUserError 

Request ID Activity ID Service Version Cluster URI 
I do query the activity log using this API endpoint https://api.powerbi.com/v1.0/myorg/admin/activityevents?  but it does not show up there. 

The experience is that not all pages of a report fail, that is quite random. as well as that even after a few minutes the report might refresh perfectly. And the error is comming up infrequently it might take a week to show up again. 
I do get a sense that it is happening more with reports that have many pages with up to 20 visuals per page. 

So that results in 3 questions
- Is this a server problem, I notice it takes place around GMT 15:00 to GMT 17:00 these are peak hours for a server hosting both European countries and Eastcoast America. for which Europe is wrapping up the day and USA states are starting their workday. So should I complain to MS and create a support ticket? 
- Is there a way to find errors like these in the API that exports activities since that would make me more aware of the size of this issue. 
- What optimizations are needed for this report, do I need to reduce the amount of visuals? Or should I focus on the datamodel. Can I limit the number of visuals that refresh at a single time? 
I read for instance this article https://learn.microsoft.com/en-us/power-bi/visuals/power-bi-data-points?source=recommendations and this article https://blog.crossjoin.co.uk/2024/06/23/power-bi-semantic-model-memory-errors-part-4-the-query-memor.... that helps to understand the concepts. But for me if the visual is inconsequent and loads on multiple days and fails on other days while data is steady makes no sense. More over, I am perfectly fine if the visuals load in sequence and not paralel. each individual visual should be able to load. Give it some time, but why fail. 

I am interested to hear both if you also have seen this error, and any suggestions are apreciated

[SofBL]

The final word on this is that on this site

https://www.microsoft.com/licensing/docs/view/service-level-agreements-sla-for-online-services

all online cloud services are having a described SLA.

For Fabric in specific that states *Downtime Minutes does not include any period of time where the inability to use Fabric features within a given Capacity was due to Microsoft Fabric throttling policy. 

This can be read as when you cannot use your reports because of errors related to throtling on a shared capacity (pro workspaces). Microsoft does not account this to services being unavailable. they are "degraded". 

I asked both Sales and Support for a reference on statistics or metrics about how often this occurs. Their flows in directions like, can you optimize your report, we advise you to use a premium capacity if you need asurance of availability. Bypassing mainly the question. If you then push forward you are offered an conference call or a sales call. And they simply inform you that this logging information is not available.

this leaves us with little options. We can either use RPA and test tools to get a glimps, technicaly this will be just sampling and will burden the service. Not a prefered route. 
the other option is the refresh log. that does show the memory throttling error as well and then if that occurs also assume reports failing. 

[V-yubandi-msft]

Thank you for sharing your findings with the community.  If you have more questions or need clarification, please feel free to ask. Your contribution is valued.

[krishnakanth240 · Accepted solution]

Hi @SofBL 

1.Server issue or your report

Mostly shared capacity.Complex reports increase the probability.Yes, you can raise a ticket if it is frequent and Microsoft will usually confirm capacity pressure.

2.Can you track this via API

Activity log does not capture query or memory throttling. So you can use Performance Analyzer in Desktop. Log Analytics if Premium. In Pro there is no reliable one for this specific error to my knowledge.

3.What to optimize

Reducing visuals per page. 20 is high, you can aim less than 10. Optimize the measures and also Data Model with star schema, removing unused columns. Using aggregations or summarized tables and limit the slicers.

Even if a visual works, it can fail under memory pressure because queries run in parallel and exceed per query memory limits.