Incremental Refresh - First scheduled refresh duration and cached data usage

[VMariapp]

I configured Incremental Refresh (6 months + 1 day) for my dataset connected to an Oracle source.

Here’s what I did:
*Loaded 6 months of historical data in Power BI Desktop (it took ~15 hours due to large volume).
*Enabled the Incremental Refresh policy in Desktop after the data was loaded.
*Published the dataset to the Power BI Service (Premium workspace).
*Did not trigger a manual refresh — instead, I set up a scheduled refresh.
Now, the scheduled refresh has been running for 3 hours, and I’m concerned it may timeout after 5 hours.
My main questions:
1. Does the first scheduled refresh in Service use the cached 6-month data from the PBIX, or does it query the entire dataset again from Oracle?
2. Since Incremental Refresh was already enabled before publishing, will it only refresh the 1-day partition or attempt a full reload on the first run?
3. Is it normal for the first refresh to take this long even when data is already present in the PBIX file?
Any insights or best practices on handling the first IR refresh with large Oracle datasets would be greatly appreciated

Thanks,
Vidhya

[v-tsaipranay]

Hi @VMariapp ,

We haven’t received an update from you in some time. Could you please let us know if the issue has been resolved? Also thank you @Poojara_D12 for your response.
If you still require support, please let us know, we are happy to assist you.

[Poojara_D12]

Hi @VMariapp 

When you publish a Power BI dataset with Incremental Refresh (IR) enabled, it’s important to understand how the first refresh in the Power BI Service behaves. Even though you already loaded six months of data in Power BI Desktop and enabled IR before publishing, the Service doesn’t directly use the cached data stored in your PBIX file. Instead, during the first refresh after publishing, Power BI performs a full partition refresh to create and store the historical and incremental partitions in the Service environment (essentially rebuilding the dataset structure for incremental logic). This means it will query all six months (the historical range) from Oracle again — not just the new one-day partition. As a result, the first refresh often takes nearly as long as your initial load in Desktop, and it’s completely normal for it to run for several hours on large datasets. Once this initial refresh is successfully completed, Power BI will maintain the partition structure, and subsequent refreshes will only query the incremental (1-day) range, making them much faster.

For best performance with Oracle sources, it’s recommended to ensure that your partitioning column (used for incremental logic) is indexed and filterable so Power BI can efficiently apply query folding. Also, consider performing the first refresh during off-peak hours or using Power BI Premium capacity with higher resource allocation to prevent timeouts. In summary: yes, the first scheduled refresh reprocesses all partitions, yes it’s normal for it to take several hours, and future refreshes will be much quicker once the incremental structure is established.

[v-tsaipranay]

Hi @VMariapp ,

Thank you for reaching out to the Microsoft fabric community forum.

Could you please let us know if the issue has been resolved? I wanted to check if you had the opportunity to review the information provided by @blopez11   . If you still require support, please let us know, we are happy to assist you.

Thank you.

[blopez11 · Accepted solution]

1. No.  2. No. it will do full refresh  3. Yes.

Incremental refresh doesn't work they way you are trying to use it.

For incremental refresh,

- You configure the incremental refresh in the desktop (creating range parameters, using them in the table you want incrementally refreshed, etc..

- Publish to to Service.

- The first refresh in the Service creates the partitions, loads historical data, etc.  This first refresh can take a bit of time.

- Subsequent refreshes will be incremental based on the policy you created.

You can avoid the first refresh taking so long by bootstrapping the initial refresh.

Check out the Guy in a Cube video: https://youtu.be/5AWt6ijJG94?si=25aIgwS6NCUDIgEI

Regards,