Microsoft Fabric query guide

After every extract, the job writes a SQL script to ADLS:

abfss://archive@<storage>.dfs.core.windows.net/<job_id>/_catalog/fabric.sql

Two ways to use it in Fabric:

Path A: Lakehouse SQL endpoint (fastest)

The SQL is the same OPENROWSET dialect that Synapse Serverless uses.

1. Create or open a Lakehouse

In Fabric → your Workspace → + New → Lakehouse.

2. Connect the Lakehouse to your ADLS

In the Lakehouse → + New → Shortcut → ADLS Gen2. Provide:

• URL: https://<storage>.dfs.core.windows.net/archive
• Authentication: Organizational account (your Entra ID), or OAuth with a service principal.

This creates a shortcut named archive under your Lakehouse’s Files/ area.

3. Open the SQL endpoint

Top-right of the Lakehouse → SQL endpoint dropdown.

4. Download fabric.sql from ADLS and paste

az storage blob download \
  --account-name <storage> \
  --container-name archive \
  --name '<job_id>/_catalog/fabric.sql' \
  --file fabric.sql \
  --auth-mode login

Paste into the SQL endpoint editor. The script:

• Creates an external data source pointing at your storage
• Creates schemas (one per Oracle schema)
• Creates a view per table using OPENROWSET

5. Query

SELECT TOP 100 * FROM HR.EMPLOYEES;

SELECT COUNT(*) FROM HR.EMPLOYEES;

-- Cross-schema join works the same as Oracle
SELECT e.name, d.dept_name
FROM HR.EMPLOYEES e
JOIN HR.DEPARTMENTS d ON e.dept_id = d.id;

Path B: Promote to native Delta tables (recommended for repeated queries)

For tables you’ll query repeatedly, convert from view-over-Parquet to a native Delta table:

-- In the Lakehouse SQL endpoint:
CREATE TABLE HR.EMPLOYEES_DELTA AS
SELECT * FROM HR.EMPLOYEES;

Or use a Fabric Notebook:

# PySpark notebook in your Fabric workspace
df = spark.read.parquet(
  "abfss://<workspace-id>@onelake.dfs.fabric.microsoft.com/<lakehouse-id>/Files/archive/<job_id>/HR/EMPLOYEES/"
)
df.write.format("delta").mode("overwrite").saveAsTable("HR_EMPLOYEES")

Delta tables give you:

• Faster queries (data co-located with compute)
• ACID transactions for incremental updates
• Time-travel queries
• Native integration with Power BI / DirectLake

Cost / performance notes

For most archive use cases, Path A is enough — you’re querying old data infrequently and don’t want to duplicate storage. Promote to Delta only for the hot tables that get repeated access.

Compared to Oracle restore

Most customers use Fabric/Synapse query for 95% of access and only restore to Oracle when an application needs the data back.