Higher than expected CU consumption by mirrored databases

[dlauth33]

I've seen some strange behavior over the last 10 days or so where my overall CU% has slowly been creeping up without any code changes or data volume increases.  I have an F128 capacity with 7-8 workspaces.  The change in CU usage seems mostly to be confined to a single workspace.  It has a central data warehouse with multiple mirrored databases and data pipelines.  I have a daily ETL that runs once a day that takes a little over an hour to complete. (pulling data from other databases within Azure as well as some data from the mirrored dbs using data pipelines)  The rest of the day the workspace mostly sits idle with very little querying being done against it or the mirrored db's. Looking at the Fabric Capacity Metrics charts, it appears that at or around April 27th, the warehouse CU jumped and continued on a steady increase until yesterday when I restarted replication on all the mirrored databases in the workspace.  Since then it's been on a mostly steady decline with no other changes.  Wondering if anyone else has seen this.   Also curious to know if this lines up up with a monthly update being applied to our capacity but I suppose that's a question for support.

I pivoted the Capacity Metric data for my workspace for the mirrored db's and found that they were consuming 25% of my total CU.  After restarting replication yesterday, it's back down to 1.5% today which is in line with the CU usage % back on April 27.  
4/23:   4%
4/27:   4.1%
4/29:  19.75%
5/5:    25.8%
5/6:      1.5%

[v-tsaipranay · Accepted solution]

Hi @dlauth33 ,

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

The increase from 4.1% to 25.8% between April 27 and May 5, 2025, was likely due to a replication backlog or synchronization issue with your mirrored databases, leading to elevated background activity. Restarting replication on May 6 cleared this backlog, reducing usage to 1.5%.

This issue isn’t connected to the April 2025 Fabric update as noted in Fabric April 2025 Feature Summary | Microsoft Fabric Blog | Microsoft Fabric, but may have been triggered by replication lag or source database update patterns. To prevent future spikes:

• Monitor replication health using the Fabric Capacity Metrics app for early detection of lag(What is the Microsoft Fabric Capacity Metrics app? - Microsoft Fabric | Microsoft Learn)
• Schedule weekly replication restarts during low-usage periods
• Set CU usage alerts (e.g., at 10%) to catch spikes early. (What is the Microsoft Fabric Capacity Metrics app? - Microsoft Fabric | Microsoft Learn)

I hope this will help you understand better, if you need any further assistance, feel free to reach out.

If this post helps, then please give us Kudos and consider Accept it as a solution to help the other members find it more quickly.

Thankyou.