Forum Discussion
Helper ISemantic Model isn't showing all data in dataflow
Hello!!
I'm really hoping someone can help me..
I've created a dataflow that contains 36 tables and feeds into numerous reports. I've set up a data pipeline that refreshes the dataflow and upon success refreshes the semantic models for my reports.
The data flow is refreshing correctly but I'm continuously running into an issue where 3-4 of my tables in the dataflow aren't sending any, or all, of their information into the semantic models causing reports to break.
I've tried opening the dataflow and refreshing the tables, I've tried making sure the dataflow saves correctly, and I've tried recreating the pipeline relationships.
Can someone please help as I feel like I am losing my mind...
I need to be able to make edits to the dataflow on occasion without worrying about validating the reports every time.
This usually happens due to metadata or schema sync issues between the dataflow and the semantic model, not because the dataflow refresh itself is failing. Basically, Schema changes (column added/removed/renamed) in the dataflow tables that the semantic model hasn’t picked up. Or Semantic model refresh using cached metadata after dataflow edits. Could you follow the steps below and try again?
Open the semantic model → Refresh → “Refresh now” (manual, not pipeline) to force metadata sync
In the semantic model, Edit tables → Refresh preview for the affected tables
If schema changed, rebind the table (remove + re-add from dataflow)
Add a delay step between dataflow refresh and semantic model refresh in the pipeline
3 Replies
- Poojara_D12
Super User
As per my knoweledge, this is a known Fabric behavior, not a mistake on your side. Dataflows can refresh successfully while schema or partition metadata doesn’t fully propagate to the semantic model, which is why a few tables randomly appear empty or partially loaded. Pipelines only enforce refresh order, not metadata consistency, so this can happen especially after dataflow edits or with larger tables. If this is becoming frequent, that’s the signal to switch to Lakehouse - semantic model: do it when you have many tables, frequent schema changes, large volumes, or need predictable refresh behavior — Lakehouse-backed models are far more stable and transparent. To avoid users hitting broken reports, add a simple validation step: after refresh, run a check (Notebook, SQL endpoint, or semantic model DMV) to confirm row counts > 0 for critical tables and fail or alert if something is missing. This gives you an early warning instead of discovering issues through broken visuals.
Winter_king94I'm unfortunately not able to try this out as my company doesn't currently allow us to use Lakehouse - semantic model. Hopefully, in the future I can use this! Thanks for confirming this wasn't a me issue!
- cengizhanarslan
This usually happens due to metadata or schema sync issues between the dataflow and the semantic model, not because the dataflow refresh itself is failing. Basically, Schema changes (column added/removed/renamed) in the dataflow tables that the semantic model hasn’t picked up. Or Semantic model refresh using cached metadata after dataflow edits. Could you follow the steps below and try again?
Open the semantic model → Refresh → “Refresh now” (manual, not pipeline) to force metadata sync
In the semantic model, Edit tables → Refresh preview for the affected tables
If schema changed, rebind the table (remove + re-add from dataflow)
Add a delay step between dataflow refresh and semantic model refresh in the pipeline
