Get certified for free when you join Fabric Data Days 2026 and dive into Fabric, Power BI, SQL, AI, and other essential data skills.
Join nowData Days is here! Join us now for 60+ days of learning, challenges, and connection. Learn more
Hello,
My company is running on a Power BI Premium P2 Capacity, and we have a small gateway cluster that is distributing our refreshes properly.
We receive batches of errors (5-15 at a time) from scheduled refreshes failing due to the number of datasets being processed concurrently.
I attempted to go to the Power BI Premium Capacity Metrics App to see said failures to investigate further, and on the Datasets report I see no blocks of failures that coincide with the batches of emails we get. Are throttled operations not counted as evictions, or are they tracked elsewhere?
Thanks for any help!
I think you want to look at the "Query Waits" tab, especially the chart at the very bottom. you can modify that to reveal the requests that waited the longest.
Now when your renderer limit is exceeded I don't know if requests get rejected immediately or if they are put in the wait queue. Maybe you can figure that out.
I appreciate the input!
I can see on that graph that over 94% of my query waits are sub 50ms, and the rest are in the 50ms-100ms bucket, so that even furthers my confusion over why batches of refreshes are getting failed but then not tracked?
I must be misunderstanding something about how the "Back end" functions as far as refreshing and failure rules goes. I will continue to hunt for the culprit 🙂
Thanks again!
Don't miss out on Data Days, June 15 through August 7. Learn Fabric, Power BI, SQL, AI and more.
Check out the May 2026 Power BI update to learn about new features.
| User | Count |
|---|---|
| 16 | |
| 10 | |
| 10 | |
| 8 | |
| 7 |
| User | Count |
|---|---|
| 40 | |
| 36 | |
| 34 | |
| 28 | |
| 20 |