A few months ago I wrote about a rare error – the “Maximum allowable memory allocation” error – that may occur when the physical machine, or node, that a semantic model is running on in the Power BI Service comes under memory pressure. Recently, someone I was working with who was doing some load testing showed me a related error:
The operation was canceled and the database was evicted to balance the CPU load on the node. Please try again later.
This error is so rare that a web search returns no results, but since at least one other colleague at Microsoft has seen it I thought it would be good to blog about it.
Put simply, this error is the CPU version of the “Maximum allowable memory allocation” error in that when CPU usage on a node gets too high then a semantic model (maybe the one that is causing the high CPU usage, maybe not, and indeed maybe there is no single culprit) is picked to be moved to a different, quieter node. When this happens any queries or refreshes that are running on the selected semantic model are cancelled and this error is returned.
It’s easy to see how load testing can trigger this error, since load testing involves running a lot of concurrent DAX queries and therefore generates high CPU usage on the semantic model. As with the “Maximum allowable memory allocation” error, if you see this error once you may just be unlucky but if you see it repeatedly in normal production use you are probably the cause, in which case enabling semantic model scale out may help but the real solution is to tune your semantic model to reduce CPU usage. How to do that is out of the scope of this blog but as with so many things in Power BI you should start by looking at the DAX in any calculated columns (if you’re getting the error during a refresh) or in your measures (if you’re getting the error when users query the semantic model), and by looking at how your data is modelled.