Hi,
From what I learned, different workspaces are treated independently and especially with different schedule, data should be refreshed well or at least the latest good partition of dataset would be passed on if the original data source has a random failure.
Now in my case, workspaces are refreshed multiple times a day and successfully, but stale data presented not lagging in a couple hours, but days (properly say data is stunt at a point of time and never get refreshed).
The red entities are not refreshed, but remained with stale data no matter what.
Case 1:
Case 2: Stuff in between a secondary dataflow in Workspace A, data loaded correctly, but still passing on stale data across workspace
Case 3: Split workspace and use Power Automate trigger to push data, problem persisted
Case 1B: A case that seemed to be treated as Crisscross even though two workspaces are separated
Any documents out there suggested that this practice is wrong? Or a Dataflow hierarchy could not be refered twice by another Dataflow hierarchy?
Kindly advise. This problem persisted since December 2022 and I saw multiple cases by others described very close to mine.
Thanks and best regards.
R.
3 Comments
Hi rdnguyen
I found some offical documents , hope it can help you !
Dataflows best practices - Power BI | Microsoft Learn
Understand and optimize dataflows refresh - Power BI | Microsoft Learn
Best Regards,
Community Support Team _ Ailsa Tao
rdnguyenHelper V
Hi Anonymous Alisa, I read through those documents / articles many times before, but I don't think I see any part clarifying Enrichment step experiencing stale data issue (by joining multiple tables) except advising developer to layering dataflows in phases.
My problem is obviously at the join that some of the joining tables may not passing on updated data, rather stale data cached at certain point of time. This is totally throw me out of water for my calculations all based on the final output which I previously confidently using all fields as updated data.
If only I added the Refresh date in, I found the issue at the root table, but hardly could I explain why think like this happened.
This is the data passed on to enrichment step from linked entity.
And this is the original query, which has new data as much frequent as refresh scheduled.
Hi
Yes I am facing the same problem, the current workaround is to not rely on linked dataflows (unload them) and reference an upstream dataflow in the same workspace in the M code (so dataflows think it cannot be skipped in the refresh)
If you check the refresh history you can see the linked dataflow was skipped (assuming the data from the earlier refresh has been pushed into this linked dataflow but clearly the linked dataflow holds stale data)
The problem is linked dataflows between workspaces do NOT always hold the same data and if this was the case, then the orchestration of refreshes would work correctly.
