PBI Service Refresh Issues -Unable to cancel refresh, refreshes stall- Affects multiple capacities

[JSanders28]

Anyone else having issues cancelling queries? Or seeing stalled queries in Power BI service?

We have 4 premium capacities, and for the past 1.5 months we have issues where semantic models will suddently stall on refresh. Yes we have a ticket with Microsoft.  Curious about other peoples experiences. 

• Refreshes that take 3-5 minutes suddenly stall and  get stuck "Refreshing" until they timeout after 5 hours.
• We are also unable to cancel these long running queries. (It gets stuck on "Canceling" and never cancels.
• It's happening on multiple capacities and gateways.
• We verified it's not the datasource.
  • No queries running on datasource as PBI is refreshing
  • Also went has far as to refresh the semantic models in Power BI desktop directly on the gateway server itself, while it's refreshing.  Performance of refreshing on gateway in Power BI is desktop is completly fine and normal and able to do that within 3-5 minutes (as usual).  
• The semantic models it happens on are different every time.

Appears to be product related. We're in US East Region.

[bkundrat]

Did you ever resolve this issue? We've just started having this issue and while I was researching whether others have had this experience, I came across your post. 

I'd be interested to know your status. Thanks!

[JSanders28]

Hello @bkundrat  we had the issue happen again yesteerday.  We now have an open ticket with Microsoft support and opened a new post.  We had the issue across multiple Power BI premium capacities and gateways.  Also opened a new idea/topic for it.  If you wouldn't mind upvoting the idea, that'd be great.

 https://community.fabric.microsoft.com/t5/Issues/Microsoft-Issue-Refreshes-Getting-Stuck-unable-to-c... 

[lbendlin]

We also continue to see that. The "problem" is that even with a Prio A ticket the Microsoft team/partner will successfully drag their feet until the "natural" 18000 second timeout makes the issue go away.  Hard to properly troubleshoot.

[JSanders28]

That's been our experience as well.  However, this time we got our account rep involved, and we've got it to the product team now, and they are looking into it.  So far they've confirmed that the issue is on the PBI service side, and seems to freeze up on the last part of saving the semantic model back to the service (after the data has been pulled from the source). 

One workaround that was suggested, if a dataset is stuck canceling, you can trigger an Enhanced API refresh.  (not the just semantic model refresh via API, but an enhanced API refresh and it will refresh while the other one is still canceling).  

[lbendlin]

Sorry to hear that.  Have you tried moving affected workspaces across capacities?