Skip to main content

Demo on Databricks Free Edition

You can try the complete MDMongoose — app, pipelines, jobs and golden records — on Databricks Free Edition, at no cost and without a credit card. The setup takes about 20 minutes, most of it waiting for the first deployment and the init job. You need no local tools: everything happens in the browser.

What you get​

The demo seeds realistic, invented data that behaves like real sources do — the same company spelled differently per system, missing VAT IDs, typos and stale values:

BranchContent
Customer1 000 companies in 7 countries from three sources: SAP kna1, Salesforce account and a webshop, matched into golden records
Supplier200 companies from SAP lfa1 and Coupa
Employee300 people from Workday and Active Directory
Product50 products in a six-level hierarchy, entered by hand
Reference DataCountries and the source records as loaded

1. Create a Free Edition account​

  1. Go to databricks.com/learn/free-edition and sign up with an e-mail address, Google or Microsoft.
  2. You get one workspace with Unity Catalog. It already has the catalog workspace and a SQL warehouse called Serverless Starter Warehouse — exactly what MDMongoose uses by default, so nothing needs to be configured.

2. Add the repository as a Git folder​

  1. In the workspace, open Workspace in the left menu, then your home folder.
  2. Click Create → Git folder.
  3. Paste the URL of the MDMongoose repository (see Resources), keep the provider GitHub, and click Create Git folder. A public repository needs no Git credentials.

3. Deploy the bundle​

  1. In the Git folder, open databricks.yml.
  2. Click the Deployments icon (the rocket) on the left of the editor.
  3. Select the target free (the default) and click Deploy. Confirm the resources it lists: the app, the jobs mdm_init, mdm_load, mdm_action and mdm_refresh, the pipelines mdm_load and mdm_export, and the schemas mdm_core, mdm_staging and mdm_export.
  4. Wait for Deployment succeeded (one to three minutes).

4. Run mdm_init​

  1. In the same Deployments panel, under Bundle resources, find the job mdm_init and click Run (▶).
  2. Follow the run under Jobs & Pipelines. On Free Edition it usually takes 10–20 minutes. It:
    • creates the MDMNodes table and the system definitions;
    • creates the seven demo source tables in mdm_staging with change data feed;
    • loads them through the mdm_load pipeline and builds the golden records for Customer, Supplier and Employee;
    • grants the app's service principal access, and runs the mdm_export pipeline once.
If the grant step reports a problem

On Free Edition you are the workspace admin, so the grants normally succeed. If the run output says it could not grant access, open sql/grants.sql from the Git folder in the SQL editor and run it once.

5. Start the app​

  1. Go to Compute → Apps and open mdm.
  2. Click Start if it is stopped, and wait until it is Running. The first start builds the UI and takes a few minutes.
  3. Click the app's URL. Sign in with your Databricks account when asked, and allow the app to act on your behalf.

You land on MDM, the root of the hierarchy. As the workspace admin, you are an MDMongoose admin too.

6. Take the tour​

Try thisWhereRead more
Open a golden record and its three source recordsMaster Data › Customer → any recordMaster data records
Explode the customers in the grid, sort, filter, pasteMaster Data › Customer → Edit ChildrenEdit Children
Run a dry run of the matchingMaster Data › Customer → Refresh Golden RecordsMatching
Change a survivorship rule and simulate itModel › Rules → Customer survivorship v2 → Edit RuleSurvivorship rules
Walk a six-level product hierarchyMaster Data › ProductModifying data
Load changes from a sourceReference Data › SAP ECC › kna1 → Load ChildsImporting data
Export a branch to Excelany node → ExportExporting to CSV and Excel
Query an exportSQL editor: SELECT * FROM workspace.mdm_export.customer_golden_recordsMaterialized views
Check what a group may doheader → View as UC groupSecurity

To see incremental loading at work, change a source row in the SQL editor and load it:

UPDATE workspace.mdm_staging.src_sap_kna1
SET TELF1 = '+49 89 5550111'
WHERE KUNNR = '0000100234'; -- ACME GmbH

Then click Load Childs on Reference Data › SAP ECC › kna1: only the changed row is merged. Because the table has Run Golden After Load, the golden record ACME GmbH is refreshed too — its Phone uses Most Recent, so the new number wins.

Living within the Free Edition limits​

Free Edition is generous enough for the demo, but it has limits. What they mean here:

LimitWhat you noticeWhat to do
Apps stop 24 hours after they were started or deployedThe app URL shows that the app is stoppedStart it again from Compute → Apps. All data is in Delta; nothing is lost.
One small SQL warehouseThe first screen after a pause takes a few seconds while the warehouse startsNothing — later clicks are fast; the app caches what you visited.
Daily compute quotaJobs and the warehouse stop for the rest of the day when it is used upKeep the hourly mdm_refresh job paused (it is by default); avoid full reloads in a loop.
One active pipeline per type, 5 concurrent job tasksA load started while an export runs waits or failsRun one load or refresh at a time.
Serverless jobs start in about a minuteLoad Childs and Refresh Golden Records show a run dialog for a whileSaves in the UI are direct and do not use jobs.

Starting over or cleaning up​

  • Reset the demo: drop the schemas and deploy again.

    DROP SCHEMA workspace.mdm_core CASCADE;
    DROP SCHEMA workspace.mdm_staging CASCADE;
    DROP SCHEMA workspace.mdm_export CASCADE;

    Then Deploy and run mdm_init again.

  • Remove everything: in the Deployments panel choose Destroy (or databricks bundle destroy), then drop the schemas as above and delete the Git folder.

Next steps​

  • Load your own data: Importing data.
  • Install on a paid workspace or without demo data: Deployment — set demo to "false" for a clean install.