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

Lakehouse Inspector — from install to a profiled medallion

Install app-lakehouse-inspector and get a bronze / silver / gold medallion lakehouse on ADLS Gen2 + Delta, seeded with the retail-sales reference star schema, plus a companion notebook that profiles every tier. It is the fastest way to see a fully formed lakehouse rather than an empty file browser. ~20 minutes.

What you end up with

Two items: Retail Sales Medallion Lakehouse (10 seeded Delta tables + 10 folders + 1 shortcut) and Lakehouse Inspector Walkthrough (notebook). Azure-native throughout — ADLS Gen2 for storage, Synapse serverless for SQL, Spark for compute. No OneLake, no Fabric capacity.

Prerequisites

You need Why If it is missing
A Loom workspace Apps install into a workspace Tutorial 01 — First workspace
LOOM_ADLS_ACCOUNT The lakehouse provisioner writes folders, seed CSVs, and Delta tables Lakehouse row shows an honest gate naming the account
Console UAMI with Storage Blob Data Contributor on that account Write the seed data Verbatim 403 plus the role to grant
A Synapse serverless SQL endpoint (recommended) Registers queryable views over the tables and the shortcut Tables still land in ADLS; the SQL views do not
A Spark compute (Synapse Spark pool or Azure ML) Run the walkthrough notebook The notebook installs but has nothing to execute on

1. Install the app

  1. Left nav → Apps → Lakehouse Inspector (/apps/app-lakehouse-inspector).
  2. Install into workspace → workspace, optional folder (e.g. Lakehouse).
  3. Deploy artifacts to live Azure services → On. Compute → Shared.
  4. Install.

2. What gets provisioned

Item Provisioner Real backend
Retail Sales Medallion Lakehouse (lakehouse) lakehouseProvisioner ADLS Gen2 folder layout; seed rows landed as CSV then loaded to managed Delta; the declared shortcut uploaded into the tenant's own storage and registered
Lakehouse Inspector Walkthrough (notebook) notebookProvisioner Synapse (LOOM_SYNAPSE_WORKSPACE) nbformat artifact, or Databricks (LOOM_DATABRICKS_HOSTNAME) SOURCE notebook under /Shared/loom-installs/…

Folder layout — it mirrors dbt exactly

Ten folders, bronze/<entity>/ → silver/<entity>/ → gold/<dim_or_fact>/:

Tier Folders What lives there
bronze sales/, customers/, products/ Raw landings. Sales partitioned by ingestion date (YYYY/MM/DD); customers and products are daily full snapshots. No transformation — materialized as views in dbt
silver sales/, customers/, products/ Cleansed. Sales type-cast and null-filtered (order_id NOT NULL, quantity > 0, unit_price >= 0), incremental on _ingested_at. Customers canonicalized to one row per customer_id + ingestion_ts. Products title-cased with list_price validated \(0–\)100K
gold dim_customer/, dim_product/, dim_date/, fact_sales/ SCD Type 2 customer and product dimensions (surrogate *_key, valid_from / valid_to / is_current), a static dim_date (1900-01-01 → 2099-12-31, date_key = YYYYMMDD), and fact_sales at order-line grain with role-playing order_date_key / ship_date_key and pre-computed extended_amount / cost_amount / margin_amount

3. Seeded data

Ten Delta tables, seeded at install:

bronze_sales, bronze_customers, bronze_products, silver_sales, silver_customers, silver_products, dim_customer, dim_product, dim_date, fact_sales.

Plus one shortcut: retail-orders-public — a 267-row retail-orders CSV that ships in the repo (samples/app-data/lakehouse-inspector/retail-orders-public.csv).

Why the shortcut is self-contained

A shortcut that points at an external URL can 404 tomorrow. This one is a repoDataset: at install the provisioner reads the real file, uploads it into your ADLS under the lakehouse's Files/_shortcuts/retail-orders-public/, registers a real internal shortcut row, and — when Synapse serverless is configured — a queryable OPENROWSET view. It demonstrates the zero-copy pattern with nothing external to reach.

The schema is the examples/fabric-e2e reference architecture: 3 dimensions + 1 fact, with the same column contracts as contracts/dim_customer.yaml, dim_product.yaml, dim_date.yaml, and fact_sales.yaml.

4. First meaningful task — profile the medallion

Open Lakehouse Inspector Walkthrough, attach a compute in the ribbon's compute picker, and run the cells. They are idempotent — run any of them in any order.

  1. List every Delta table across all three tiers — iterates spark.catalog.listTables(tier) for bronze / silver / gold and shows tier, table, and type. Expect ten rows. This is the single best "did the install actually work" check.
  2. Row counts per tier — the profiling baseline.
  3. Null counts on primary keys — the data-quality smoke test. order_id, customer_id, and product_id must be zero-null; anything else means the seed or your own load broke the contract.
  4. Sample queries against the gold star schema — fact_sales joined to its dimensions:
  5. top customers by margin,
  6. category margin percentage.
  7. Read the retail-orders-public shortcut — proves the zero-copy path resolves from Spark, not just from the file browser.

Then leave the notebook and open the lakehouse editor itself:

  • The Files tree shows raw uploads on ADLS Gen2 (including Files/_shortcuts/retail-orders-public/); the Tables tree shows managed Delta.
  • Drop your own CSV into a bronze/<entity>/ folder and use Load to Tables to infer the schema and write a managed Delta table — no DDL.
  • Query the SQL analytics endpoint with T-SQL, or read Delta from a notebook with spark.read.format('delta').

5. Verify it worked

  • Install dialog: the lakehouse row is created, and its step log lists the folder creations, the seeded table loads, and the shortcut upload + registration.
  • Notebook cell 1 returns ten tables across the three tiers.
  • Notebook cell 3 returns zero nulls on every primary key.
  • Lakehouse editor → Files shows Files/_shortcuts/retail-orders-public/ with the CSV in it.

Troubleshooting

Symptom Cause Fix
Lakehouse row = remediation LOOM_ADLS_ACCOUNT unset Set it, restart the Console revision, Retry the row
Verbatim 403 on the seed write UAMI lacks storage rights Grant Storage Blob Data Contributor on the ADLS account
Cell 1 returns fewer than ten tables Some seeds failed, or the Spark catalog is not pointed at the lakehouse databases Read the install step log for the failed loads, then re-run the install (idempotent)
Shortcut folder exists but no SQL view Synapse serverless is not configured Configure the serverless endpoint; the file-level shortcut still works from Spark
Notebook has no compute to run on No Spark pool / AML compute attached Pick one in the ribbon's compute picker; a cold Synapse Spark pool takes ~2 minutes for the first session
Notebook cell fails with a catalog error The Spark session's default filesystem container does not exist Loom sets fs.azure.createRemoteFileSystemDuringInitialization=true on Livy sessions; if you are running outside Loom, create the container first

Cleanup

Delete both items, or the workspace. The ADLS folders, Delta tables, and the uploaded shortcut file are real — remove them from storage if you want the space back.

What's next