Forum Discussion
DEV Fabric Workspace Refresh Failure
Fabric F8 Workspace - Incremental Refresh Fails with Resource Governance (Memory Exceeded) During Dataset Refresh
Hi Team,
We are experiencing a refresh issue in our Microsoft Fabric (F8) workspace while refreshing a Power BI Import semantic model that uses Incremental Refresh.
Environment
Workspace Type: Microsoft Fabric (F8)
Dataset Mode: Import
Data Source: Amazon Athena
Refresh Type: Incremental Refresh
Incremental Refresh is configured using UPDATE_DATETIME as the partition/filter column.
Issue
The dataset refresh fails with a Resource Governance error indicating insufficient memory.
Error message:
Data source error: Resource Governance: This operation was canceled because there wasn't enough memory to finish running it.
Example details:
Consumed memory: 2637 MB
Memory limit: 2637 MB
Database size before command execution: 434 MB (or approximately 449 MB)
Observations
The semantic model already exists and loads successfully in the Power BI Service.
Existing reports continue to work correctly.
Only the dataset refresh fails.
This issue occurs consistently in the Fabric workspace.
Questions
What exactly causes this Resource Governance error during refresh?
Is the memory being consumed by:
Partition processing?
Temporary tables created during refresh?
Query folding issues?
Incremental Refresh processing?
How can we identify which table or partition is consuming most of the memory?
Are there any Microsoft-recommended best practices for optimizing Incremental Refresh in Fabric to avoid this issue?
Is this expected behavior for an F8 capacity?
Any guidance or documentation would be appreciated.
Thank you.
Hi manoj_0911
This is a refresh-time memory limit issue in Fabric F8, not a report-rendering issue.
Even though the semantic model loads and reports work, the incremental refresh operation can still exceed the memory available for processing partitions.In your case, the key clue is:
- Consumed memory = 2637 MB
- Memory limit = 2637 MB
That means refresh hit the capacity guardrail exactly and was canceled.
What to do
1) Make sure query folding is preserved
For Incremental Refresh to work efficiently, Power Query must push the date filter down to Athena.
If folding breaks, Power BI may pull too much data into memory during refresh.
Action
- Verify the UPDATE_DATETIME filter is applied as early as possible in Power Query
- Avoid transformations before the filter that break folding
- Check whether Athena is actually receiving the filtered query
Microsoft docs
- Incremental refresh overview:
https://learn.microsoft.com/en-us/power-bi/connect-data/incremental-refresh-overview - Incremental refresh troubleshooting:
https://learn.microsoft.com/en-us/power-bi/connect-data/incremental-refresh-troubleshoot
2) Reduce the amount of data processed per refresh
If the refresh window is too large, or too many partitions are being reprocessed, memory usage will spike.
Action
- Reduce the refresh period if possible
- Keep only the required historical range
- Reprocess fewer partitions at a time if you are managing partitions manually
3) Remove unused and high-cardinality columns
Model width is one of the biggest causes of refresh memory pressure.
Action
- Remove columns not used in reports, relationships, or DAX
- Reduce high-cardinality text columns
- Replace long text or GUIDs with integer keys where possible
- Convert datetimes to dates if time precision is not needed
Microsoft docs
- Import modeling data reduction:
https://learn.microsoft.com/en-us/power-bi/guidance/import-modeling-data-reduction
4) Simplify Power Query transformations
Heavy M transformations increase refresh overhead and may break folding.
Action
- Move shaping work upstream if possible
- Keep Power Query steps minimal
- Filter early, remove columns early, and avoid expensive custom logic in the semantic model
5) Check capacity headroom
An F8 capacity is relatively small for import refresh workloads.
Action
- Use the Fabric Capacity Metrics App to see whether the refresh is pushing the capacity to its limit
- If the model is close to the limit, consider scaling up the capacity or reducing model size
Recommended interpretation
For Fabric F8, this behavior is not unusual when:
- the table is large,
- the refresh window is broad,
- folding is imperfect,
- or the model has many wide/high-cardinality columns.
So the likely fix is either:
- optimize the model and refresh pattern, or
- move to a larger Fabric capacity
Some more microsoft documentation links for your reference :
Incremental refresh overview:
https://learn.microsoft.com/en-us/power-bi/connect-data/incremental-refresh-overviewIncremental refresh troubleshooting:
https://learn.microsoft.com/en-us/power-bi/connect-data/incremental-refresh-troubleshootData reduction guidance for import models:
https://learn.microsoft.com/en-us/power-bi/guidance/import-modeling-data-reductionRefresh troubleshooting scenarios:
https://learn.microsoft.com/en-us/power-bi/connect-data/refresh-troubleshooting-refresh-scenarios
2 Replies
Hi manoj_0911
This is a refresh-time memory limit issue in Fabric F8, not a report-rendering issue.
Even though the semantic model loads and reports work, the incremental refresh operation can still exceed the memory available for processing partitions.In your case, the key clue is:
- Consumed memory = 2637 MB
- Memory limit = 2637 MB
That means refresh hit the capacity guardrail exactly and was canceled.
What to do
1) Make sure query folding is preserved
For Incremental Refresh to work efficiently, Power Query must push the date filter down to Athena.
If folding breaks, Power BI may pull too much data into memory during refresh.
Action
- Verify the UPDATE_DATETIME filter is applied as early as possible in Power Query
- Avoid transformations before the filter that break folding
- Check whether Athena is actually receiving the filtered query
Microsoft docs
- Incremental refresh overview:
https://learn.microsoft.com/en-us/power-bi/connect-data/incremental-refresh-overview - Incremental refresh troubleshooting:
https://learn.microsoft.com/en-us/power-bi/connect-data/incremental-refresh-troubleshoot
2) Reduce the amount of data processed per refresh
If the refresh window is too large, or too many partitions are being reprocessed, memory usage will spike.
Action
- Reduce the refresh period if possible
- Keep only the required historical range
- Reprocess fewer partitions at a time if you are managing partitions manually
3) Remove unused and high-cardinality columns
Model width is one of the biggest causes of refresh memory pressure.
Action
- Remove columns not used in reports, relationships, or DAX
- Reduce high-cardinality text columns
- Replace long text or GUIDs with integer keys where possible
- Convert datetimes to dates if time precision is not needed
Microsoft docs
- Import modeling data reduction:
https://learn.microsoft.com/en-us/power-bi/guidance/import-modeling-data-reduction
4) Simplify Power Query transformations
Heavy M transformations increase refresh overhead and may break folding.
Action
- Move shaping work upstream if possible
- Keep Power Query steps minimal
- Filter early, remove columns early, and avoid expensive custom logic in the semantic model
5) Check capacity headroom
An F8 capacity is relatively small for import refresh workloads.
Action
- Use the Fabric Capacity Metrics App to see whether the refresh is pushing the capacity to its limit
- If the model is close to the limit, consider scaling up the capacity or reducing model size
Recommended interpretation
For Fabric F8, this behavior is not unusual when:
- the table is large,
- the refresh window is broad,
- folding is imperfect,
- or the model has many wide/high-cardinality columns.
So the likely fix is either:
- optimize the model and refresh pattern, or
- move to a larger Fabric capacity
Some more microsoft documentation links for your reference :
Incremental refresh overview:
https://learn.microsoft.com/en-us/power-bi/connect-data/incremental-refresh-overviewIncremental refresh troubleshooting:
https://learn.microsoft.com/en-us/power-bi/connect-data/incremental-refresh-troubleshootData reduction guidance for import models:
https://learn.microsoft.com/en-us/power-bi/guidance/import-modeling-data-reductionRefresh troubleshooting scenarios:
https://learn.microsoft.com/en-us/power-bi/connect-data/refresh-troubleshooting-refresh-scenarios
Hi ,
Thank you for reaching out to Microsoft Fabric Community Forum, Below are the few points which can resolve your issue.
- The refresh failure occurs because the Incremental Refresh operation exceeds the memory limit available in the Fabric F8 capacity.
- The memory consumed during refresh is not only the dataset size; it also includes Existing semantic model memory, Incremental partition processing, Temporary processing tables, VertiPaq compression and optimization during refresh.
- Verify that Incremental Refresh is configured correctly Ensure RangeStart and RangeEnd filters are applied properly Confirm UPDATE_DATETIME filtering is being pushed to Amazon Athena through query folding, Ensure the refresh is not triggering a full dataset refresh.
- Optimize the data model to reduce memory usage Remove unused columns,Reduce high-cardinality columns, Avoid unnecessary calculated columns and calculated tables, Follow a star schema design.
- Analyze refresh memory usage using Power BI Service Refresh History,DAX Studio and VertiPaq Analyzer to identify large tables/columns,XMLA endpoint tools to analyze partition processing.
- Reduce refresh workload by Limiting the incremental refresh window Reducing partition size Avoiding multiple simultaneous refresh operations.
- F8 capacity has limited resources, and large or complex Import model refreshes can exceed available memory. If optimization does not resolve the issue, consider increasing the Fabric capacity size.
Thanka,
Chaithanya.
