Reports refresh failures - There's not enough memory to complete this operation.

[Anonymous]

How is memory allocated for report refreshes in Power BI Service?

I've never had any issues refreshing reports in the past, but some of the reports we haven't touched in months are now starting to encounter refresh issues due to the assigned cluster running out of memory. We haven't published any new reports or altered any existing ones either, so also wondering if Microsoft has recently changed how much memory is allocated per user/workspace?

The error doesn't happen all the time, it first started this Tuesday by affecting one report, then Wednesday 3 reports, and yesterday 6reports. We use Power BI Pro, so no Premium capacity etc, and no personal gateway.

For now I have changed so some reports are refreshed 30minutes later and hope that will fix it. But if memory is allocated per user or workspace another alternative might be to have someone else take over the reports or publish to another workspace.

[ibarrau]

Interesting. I'm know that the organization is hosted in one region but I'm not sure if it's the same vm for all the workspaces. I think it might be one workspace for each vm shared capacity. If you have the reports on the same workspace you should distribute the refresh time for sure.

Try distributing that and we can check the results after that 🙂

[Anonymous · Accepted solution]

Thanks, so it hasn't happened since last Friday, as it only happened a couple of times last week and hasn't re-occured it was likely an issue with the specific cluster which has since been fixed by Microsoft.

[ibarrau]

Hi. Can you tell us which are the data source? What's the size of the pbix file? do you have any big table with millions of rows?

Just remember that Pro refreshes has some limitations, like 1gb model size, 10gb for single table processing and 2 hours for a refresh. There is also problems with memory for data gateways.

Regards,

[Anonymous]

Thanks for the reply,

Well the issue isn't with any individual reports as the smaller ones that are affected are less than 5mb and 70k rows and takes less than 5minutes to refresh. The issue appears to be that a lot of reports simultaneously refresh, and their usage is pooled in the Microsoft cluster under the same VM. I'm not sure if Microsoft has recently changed how they distribute these resources/how much resources they provide as we've never had issues before but are now starting to encounter this issue very frequently even sometimes when almost no report is refreshing, I'm therefore starting to believe that It could also be that it's one specific cluster that is currently not running as expected, as it's always WABI-WEST-US-C-PRIMARY that encounters issues.

The question I have is if Microsoft is pooling report refresh usage under the same user, same workspace, or same organization. Because if it's based on user or workspace then we can split some reports to be owned by a different user or workspace. If it's based on Organization then I suppose we either have to distribute the refresh hours more or upgrade to Premium capacity, or think hard on which reports are actually needed.