Skip to content
CSA Loom — the Microsoft Fabric experience for Azure tenants where Fabric isn't yet available: lakehouses, warehouses, notebooks, semantic models, Activator rules, Data Agents, across Commercial, GCC, GCC-High, and DoD IL5

Tutorial 06 — Mirroring from Cosmos DB

Create a Mirrored database item that snapshots a Cosmos DB container into an ADLS Bronze layer, then explore it downstream with Weave. 30 minutes.

Prerequisites

  • Workspace from previous tutorials
  • A Cosmos DB account + database + container (or the test container in examples/fiab-quickstart/cosmos-seed/)
  • A connection (Key Vault-backed) to that Cosmos account — you can create one inline in the wizard

The Azure-native mirror engine copies to ADLS Bronze; no real Fabric mirroring is required.

How the Mirrored database editor works

You create a Mirrored database item with a guided wizard (source card → Key-Vault connection → optional table subset → Start). The engine snapshots source tables to ADLS Bronze CSV and exposes per-table metrics plus ready-to-paste OPENROWSET SQL. Start/Stop is a single /state call. There is no mirroring CLI and no per-workspace "Mirroring pane".

Steps

1. Create a Mirrored database

Left nav → Workspaces → open your workspace → New item → category Data Engineering → Mirrored database. The Mirrored database editor opens with its create wizard.

2. Choose the source

In the wizard, pick a source card. The available sources are Azure SQL Database, Azure SQL MI, Azure Database for PostgreSQL, Cosmos DB, Snowflake, SQL Server 2025, SQL Server 2016-2022, and Open mirroring. Choose Azure Cosmos DB.

3. Workspace, name, connection

Pick the workspace (loaded from GET /api/loom/workspaces), enter a mirror display name, and choose a connection. The connection picker uses the Loom ConnectionBuilder — select an existing Key Vault-backed credential or create one inline (it POSTs to /api/loom/connections).

4. (Optional) Pick a table/container subset

The wizard can list source tables/containers (GET /api/items/mirrored-database/new/tables?…). Leave the selection empty to mirror everything, or check specific containers.

5. Create

Review the summary and click Create (POSTs to /api/items/mirrored-database). The item lands in the workspace tree.

6. Start the mirror

In the editor, click Start (POSTs { action: 'start' } to /api/items/mirrored-database/<id>/state). The engine snapshots the source to ADLS Bronze CSV at LOOM_BRONZE_URL/mirrors/<wsId>/<mirrorId>/<schema>.<table>/snapshot.csv. For Cosmos DB it reads each container with a data-plane SELECT * FROM c via the Cosmos SDK (SQL-family sources use TDS plus change-feed enablement).

7. Inspect the per-table metrics

The per-table grid refreshes from the run result: row count, bytes, last sync, truncation flag, and a copyable OPENROWSET query. Click Copy SQL for a table to get a Synapse Serverless query ready to paste into a lakehouse SQL tab or a notebook:

SELECT TOP 100 *
FROM OPENROWSET(
  BULK 'abfss://.../mirrors/<wsId>/<mirrorId>/<schema>.<table>/snapshot.csv',
  FORMAT = 'CSV', PARSER_VERSION = '2.0', HEADER_ROW = TRUE
) AS rows

8. Explore downstream (Weave)

Use the item's Weave edges to go further:

  • Explore in Notebook (POST /api/thread/mirror-to-notebook) opens a notebook prefilled with a Spark cell per table reading its abfss path.
  • Add to Lakehouse (POST /api/thread/mirror-to-lakehouse) creates a file shortcut per table into a lakehouse you choose.

From there you can transform Bronze → Silver with the Tutorial 02 pattern.

9. Stop the mirror

Click Stop (POSTs { action: 'stop' } to the /state endpoint).

The honest gap

Today the engine does a snapshot copy. Ongoing Cosmos CDC, plus the Azure-native copy runtime (ADF / Synapse Link) for Snowflake and Open mirroring, are disclosed follow-ups: choosing those source cards shows an honest gate on Submit rather than silently stubbing. See mirrored-database.

What's next

Cleanup

  • Editor: Stop the mirror, then delete the item from the workspace tree (right-click → Delete)
  • The Bronze CSV in ADLS remains; remove it manually if desired

Troubleshooting

  • Empty table list: verify the connection has read permission on the Cosmos account
  • Snapshot fails for Snowflake / Open mirroring: those are gated follow-ups — use a SQL-family or Cosmos source for now
  • Schema differs from source: the snapshot reflects the source at copy time; re-run Start to refresh