The ultimate Fabric, Power BI, SQL, and AI community-led learning event. Save €200 with code FABCOMM.
Get registeredCompete to become Power BI Data Viz World Champion! First round ends August 18th. Get started.
Hi Community,
we have refreshed a dataset, but something happened. We don't know exactly what. The SQL profiler is silent, and there's no active session on the database either. But somehow, there's an active session that blocks new refreshes. Since we want to refresh our datasets every hour, we are completely stuck. There are several tutorials on how to cancel a refresh session - like these ones:
* Disconnect Users and Sessions on Analysis Services Server | Microsoft Learn
* Cancelling a Power BI Premium Dataset Refresh (elegantbi.com)
So we cancelled a session without using <CancelAssociated>1</CancelAssociated>
And yay! - it worked:
Executing the query ...
Run complete
However, the connection apparently stays open and we cannot start a new refresh.
The session (or connection?) is still blocking new refreshes, until 5 hours later, the connection expires and new refreshes are possible again. From our research, session status = 3 means that the session is killed. But it's still there!
MSFT support claims there's nothing they can do. We will have to wait until something magical happens 5 hours later and new refreshes are possible again.
My question to the community:
Are there ways to start a new XMLA refresh even though one connection is being blocked/locked/not properly killed?
BTW: This error occurs since 2 weeks nearly once per day on different datasets, the 1,5 years before it never occured. Hmmmmm
Solved! Go to Solution.
Your current semantic model contents will be preserved but the potential changes from the refresh will be discarded.
Hi @rks ,
You can use rest API commands.
https://learn.microsoft.com/en-us/rest/api/power-bi/datasets/cancel-refresh
Thanks,
Sai Teja
If you are really desperate then move the workspace off Premium, and then back on. That sometimes helps to unstick it.
Hm, good idea @lbendlin, it seems like a restart, which always worked with Analysis Service back in the days.
The datasets are quite large, would data loss occur in these scenarios?
Your current semantic model contents will be preserved but the potential changes from the refresh will be discarded.