TLDR: Try restarting the Power BI Service. If this doesn't work, increase your SKU capacity - things won't go back to normal immediately, but it should happen sooner. Decrease your SKU capacity back to the original level when things are normal - or when you've run the higher capacity for long enough, given the extra costs!
Hi Rahul, I spent some time back and forth with Microsoft on this, having raised a severity A support request - the issue essentially took out all reporting on a customer facing app. (And I still await some unanswered questions around how the issue came to be and why it occurred given no significant change to the load on the Power BI service). But putting that aside...
In my case, MS determined the issue was due to throttling - and provided this article to explain it. It makes sense, even if (for my scenario) I don't believe throttling should have occurred in the first place. Note that I use Power BI Embedded.
From the correspondence with Microsoft and reading the article, understanding there is a thing called carry forward is probably one of the most important things to know in this situation. Essentially, a build up of over capacity workload is carried forward and gradually reduces as capacity frees up. Note that more workload (e.g. Power BI datasests getting refreshed) will likely get added to the overall workload Power BI has - while throttling is happening - and this potententially extends the length of time throttling lasts for.
In my case, I got throttled and it was over 2 days before things returned to normal.
The solution (says Microsoft) is to increase the capacity of your SKU. In my case, I roughly doubled it, from A1 SKU to the A2 SKU. But, this made no good difference - it still took 7 hours after this doubling of SKU capacity for things to return to normal. In hindsight, I probably should have upped the SKU to an A6 or higher, to clear the carry forward, get the reporting working again, then reduce back to an A1.
Note: The usual utilisation - CPU (Gen2) - I see via metrics in Azure is 15-25%. Throttling started about the same time as utilisation hit 100%, but thottling continued while utilisation dropped to 75% and continued for many hours while at under 20%...so current CPU (Gen2) usilisation does not correlate with whether you are being throttled or not. You determine when you are throttled by using the Microsoft Fabric Capacity Metrics app which you get via Apps in the Power BI Service.
Microsoft also offered a workaround, which was to stop and start the Power Bi (embedded in my case) resource in Azure. This is to clear the carry forward - which would still get processed, and with Microsoft saying they would not charge for this boost of capacity, since Azure Fabric is relatively new (or something to that effect). In my case, the restart did nothing, so it might not even be a valid workaround anymore.
Hope all this helps.
