SQLink quick start
Screenshot note: Screenshots are illustrative. Controls or labels may remain untranslated in the demo release, and pages without a faithful English capture retain the source illustration. Control names and layouts can vary by release.
Start with a specific business question, then initialize resources, harvest metadata, build lineage, verify accuracy, and analyze results. Later steps depend on the systems, namespaces, and metadata established earlier.
Try online parsing
Use SQL Lineage to understand how code maps to objects. This parses SQL text without executing it against a database or replacing formal metadata harvesting.
- Open SQL Lineage from the left navigation.
- Select Oracle and enter the following fictional example.
- Leave forced metadata association disabled and click Parse Lineage.
- Check the flow from
demo_input.customerstodemo_output.customer_names. - Expand both tables' two fields and check
customer_idandcustomer_name.
INSERT INTO demo_output.customer_names
(customer_id, customer_name)
SELECT customer_id, customer_name
FROM demo_input.customers;The same workflow and result view are illustrated in online SQL lineage. Line breaks in the SQL aid readability. The objects exist in that result; the illustration does not indicate that real database metadata has been harvested. Verify source, target, arrow direction, and field names. If parsing fails, check engine and complete SQL.
For temporary SQL analysis, continue with that guide. Administrators maintaining real system lineage should follow the lifecycle below.
Implementation stages
| Stage | Work | Completion evidence |
|---|---|---|
| Define the scenario | Goals, systems, plan | Question, scope, and owner are explicit |
| Initialize | Users, organizations, systems, sources, tags | Required base objects exist |
| Harvest metadata | Inventory, namespaces, harvesting | Target metadata is in the lineage space |
| Build lineage | Import files, preprocess, parse | Cross-system relationships are queryable |
| Validate | Parsing, coverage, structural quality | Key paths meet acceptance criteria |
| Analyze | Catalog, map, and scenarios | Results answer the original question |
Step 1: Define the business scenario
Specify the object, direction, output, and reviewer. Questions might concern downstream tasks and reports affected by a change, the origin of a quality issue, systems traversed by a business flow, or nodes without matching metadata.
Inventory source systems, processing systems, warehouses, reports, and file exchanges. Record technology, environment, owner, sources, harvesting methods, and available SQL or ETL files. Prioritize onboarding and begin with a representative path that can be checked manually.
Step 2: Initialize the system
Confirm accounts, organizations, and functional permissions for administrators, consumers, and developers. Menu visibility does not grant access to every object.
Register application systems to provide business ownership for sources and high-level graph objects. Names and abbreviations identify systems; referenced systems cannot be deleted directly. Create reusable source connections and test after credential changes. Connectivity does not verify harvesting scope. Prepare consistent tags for temporary tables, core paths, or responsibility domains.
Step 3: Harvest metadata
Inventory databases, reports, files, and custom metamodel resources. Check scope, service account permissions, and refresh frequency.
Namespaces provide top-level identity. Separate development, test, production, and systems with colliding paths. Set a default schema when SQL omits it.
- Create a database, report, or file harvesting task.
- Select application system, source, and scope.
- Assign the correct namespace.
- Configure supported sampling, name enrichment, and blacklist options.
- Save, execute, and monitor the task.
- Verify Database Resources or Report Resources in the lineage space.
See resource synchronization and namespace mappings.
Step 4: Build lineage
| Source | Method | Key settings |
|---|---|---|
| SQL scripts | Script task | Engine, namespace, default schema |
| ETL files | Tool-specific ETL task | Source and target connections, mappings, preprocessing |
| Custom Excel relationships | Download, complete, and import a template | Object paths and namespaces |
| Database objects | Synchronize then parse automatically or manually | Engine and metadata constraints |
| Reports | Synchronize and parse preparation SQL | Connection namespaces |
| OpenLineage | Consume standard events | Event source and job state |
Check variables and every connection mapping before parsing. An unmapped connection can cause failure.
Choose the type matching the code source, then verify engine, namespace, and import mode. The exact task form depends on the deployed SQLink release; see parsing tasks.
Step 5: Validate accuracy
Parsing success: inspect successful, partially successful, parsing, failed, and unparsed files. Open highlighted failures and details for partial results.
Metadata coverage: compare lineage nodes against database and report metadata in the same namespace. Investigate harvesting scope, names, and namespace when nodes cannot match real tables or fields.
Structural quality: investigate isolated tables, cycles, breaks, and field consistency using source code, schedules, and business processes.
Step 6: Analyze lineage
Use Data Catalog for precise object lookup, Lineage Map for global flows, general scenarios for persistent monitoring and similar-path comparisons, Impact Analysis for changes, Quality Analysis for graph issues, and SQL Lineage for immediate code inspection.
Completion checks
- Systems, sources, and namespaces match the plan.
- Metadata is synchronized and key objects are searchable.
- Failed and partial files are resolved or recorded.
- Owners verify sample endpoints, intermediate nodes, and direction.
- Users access paths and reports within permissions.
- Analysis answers the original business question.