How to Avoid Connection Timeouts During Long-Running Operations in Microsoft Fabric

Question

How can I prevent connection timeouts during long-running Refresh or Integrate operations when replicating to a SQL database in Microsoft Fabric?

Environment

• HVR 6
• Target: SQL database in Microsoft Fabric

Answer

Microsoft Fabric terminates database connections that remain open for 35 minutes or longer. This can interrupt long-running Refresh and Integrate operations — particularly bulk loads or full-table refreshes on large datasets — before they complete, and typically results in the following error:

F_JD20F2: DBMS error [TCP Provider: Error code 0x68. Communication link failure]

To avoid this, provision an HVR agent (or hub) on an Azure Virtual Machine (VM) in the same region as your Microsoft Fabric environment, or in close network proximity to it. Running HVR operations from an Azure VM provides a more stable, low-latency network path to Microsoft Fabric and significantly reduces the risk of connection drops during long-running operations.

HVR can be deployed on the Azure VM in one of the following ways:

• Agent mode: Deploy HVR as an agent on the Azure VM and control it from your existing on-premises HVR hub. This lets you keep your current hub infrastructure while offloading Fabric-bound operations to a cloud-hosted agent with a more reliable network path.
• Hub-only mode: Deploy HVR directly on the Azure VM, with no dependency on an on-premises hub. This is a self-contained option where the hub manages all operations from within Azure, offering the most straightforward path to stable connectivity with Microsoft Fabric.

For the full list of requirements for using SQL database in Microsoft Fabric as a target, see SQL database in Microsoft Fabric as Target.