What causes Fabric Warehouse Error 24706?

Two or more concurrent transactions attempting to UPDATE, DELETE, MERGE, or TRUNCATE the same table simultaneously. Long-running transactions holding snapshot isolation while other transactions modify overlapping rows. High-concurrency ETL pipelines writing to shared staging or fact tables without coordination. Retry logic that re-executes transactions without proper backoff, amplifying conflict frequency

How do I fix Fabric Warehouse Error 24706?

Step 1: Identify the conflicting transactions by reviewing query execution logs in the Fabric Warehouse monitoring view to find overlapping write operations on the affected table.. Step 2: Implement retry logic with exponential backoff in your pipeline or application code — snapshot isolation conflicts are transient and retrying the aborted transaction is the recommended resolution.. Step 3: Redesign concurrent write patterns to avoid overlapping DML on the same table; consider serializing writes to shared tables or partitioning workloads so transactions target distinct row sets.. Step 4: If using MERGE or TRUNCATE in parallel pipelines, introduce explicit sequencing (e.g., pipeline dependencies or staging table swaps) to eliminate simultaneous access to the same target table.. Step 5: Monitor transaction duration and reduce the scope of long-running transactions to minimize the window during which conflicts can occur.

High severitydata formatMicrosoft Fabric →

Power BI Refresh Error:
Fabric Warehouse Error 24706

What does this error mean?

A snapshot isolation transaction was aborted because another concurrent transaction modified or deleted rows in the same table. This write-write conflict prevents the operation from completing safely.

Common causes

1Two or more concurrent transactions attempting to UPDATE, DELETE, MERGE, or TRUNCATE the same table simultaneously
2Long-running transactions holding snapshot isolation while other transactions modify overlapping rows
3High-concurrency ETL pipelines writing to shared staging or fact tables without coordination
4Retry logic that re-executes transactions without proper backoff, amplifying conflict frequency

How to fix it

1Step 1: Identify the conflicting transactions by reviewing query execution logs in the Fabric Warehouse monitoring view to find overlapping write operations on the affected table.
2Step 2: Implement retry logic with exponential backoff in your pipeline or application code — snapshot isolation conflicts are transient and retrying the aborted transaction is the recommended resolution.
3Step 3: Redesign concurrent write patterns to avoid overlapping DML on the same table; consider serializing writes to shared tables or partitioning workloads so transactions target distinct row sets.
4Step 4: If using MERGE or TRUNCATE in parallel pipelines, introduce explicit sequencing (e.g., pipeline dependencies or staging table swaps) to eliminate simultaneous access to the same target table.
5Step 5: Monitor transaction duration and reduce the scope of long-running transactions to minimize the window during which conflicts can occur.

Frequently asked questions

Is Error 24706 the same as Error 24556, and do I fix them the same way?

They are closely related — both are snapshot isolation update conflicts — but 24706 specifically occurs when you attempt to UPDATE, DELETE, MERGE, or TRUNCATE a table that a concurrent transaction has already modified. Error 24556 covers read-path conflicts where rows were deleted or updated. The fix strategy (retry with backoff, reduce concurrency) is the same for both.

Will simply retrying always resolve this error?

A single retry often succeeds for low-concurrency scenarios, but if your pipelines structurally overlap on the same table, retries will keep failing. Use retry as a short-term fix while you investigate and redesign the write concurrency pattern.