Merk
Tilgang til denne siden krever autorisasjon. Du kan prøve å logge på eller endre kataloger.
Tilgang til denne siden krever autorisasjon. Du kan prøve å endre kataloger.
This article answers the most common questions about Oracle to Azure PostgreSQL schema conversion in the Visual Studio Code PostgreSQL extension. The feature uses Microsoft Foundry to convert Oracle database schemas into Azure PostgreSQL DDL.
Which Azure database targets are supported?
You can use Azure Database for PostgreSQL flexible server or Azure HorizonDB for both the schema validation database and deployment destination. Match the PostgreSQL major version and required extensions between the schema validation database and deployment destination. See the schema validation database requirements.
Is the converted schema production-ready?
Not automatically. The tool generates Azure PostgreSQL DDL for review and validation. A Converted status or successful compilation doesn't establish a complete implementation, equivalent runtime behavior, or production readiness. For the recommended validation flow, see Best practices for Oracle to Azure PostgreSQL schema conversion.
What does the conversion tool do?
The tool reads metadata from the selected Oracle schemas and generates Azure PostgreSQL artifacts for review and deployment to either supported Azure service:
- Azure PostgreSQL DDL files. Object files in
postgres_ddlfor tables, views, functions, procedures, triggers, types, sequences, and indexes. - Conversion reports. Structured summaries of what was converted, what was skipped, and any items that need review.
For a full overview of inputs, outputs, and the conversion pipeline, see What is Oracle to Azure PostgreSQL schema conversion?
Is there an extra cost for the schema conversion feature?
The schema conversion feature has no separate extension charge. Budget for the services you use:
- Microsoft Foundry usage: the AI tokens consumed during conversion. For pricing, see Microsoft Foundry pricing.
- Azure Database for PostgreSQL flexible server: the target server (and any server hosting a schema validation database). For pricing, see Azure Database for PostgreSQL pricing.
- Azure HorizonDB: the resources that host the schema validation database and deployment destination if you use this service. Review pricing for your selected cluster configuration.
- GitHub Copilot: the subscription and any usage charges associated with Ask and Fix, according to your plan and organizational settings. These charges are separate from Foundry charges. See GitHub Copilot billing.
You might incur charges for other resources you choose, such as a conversion virtual machine. The tutorial prerequisites list the Copilot plans for this workflow.
How do I get started?
Follow the end-to-end walkthrough in the Tutorial: Convert Oracle schemas to Azure PostgreSQL. The tutorial covers installing the extension, connecting to the Oracle source and selected Azure target, configuring Microsoft Foundry, and running the first conversion.
How does the conversion work?
The Visual Studio Code PostgreSQL extension generates PostgreSQL definitions from an Oracle schema by using a Microsoft Foundry model deployment. It processes conversion in batches, attempts compilation in the schema validation database, and summarizes the results in a customer summary report.
Is my schema or data sent to the AI? Is anything retained?
Distinguish automated conversion, Copilot assistance, and the files kept in your workspace:
| Workflow | Data to consider |
|---|---|
| Foundry conversion | Extracted schema definitions and metadata, including names, routine bodies, views, and triggers, are sent to the configured model. Automated extraction reads Oracle catalogs, not application-table rows. Definitions can still contain sensitive literals or comments. |
| Copilot Ask and Fix | Object details and context supplied to Copilot are handled under your Copilot plan, settings, and service policies. Review attached files, prompts, and proposed tool actions; don't assume the metadata-only extraction boundary applies to everything you share in Chat. |
| Local project artifacts | Generated SQL, reports, and logs remain in the schema conversion workspace until you manage them. Review their contents before sharing and apply your organization's access and retention policies. |
Don't assume zero retention. Foundry retention depends on the APIs and features used and the applicable abuse-monitoring configuration. See Data, privacy, and security for Foundry Models. Copilot has separate data-handling policies; see the GitHub Trust Center. Review both services' requirements before submitting sensitive schema definitions.
Where does the model run?
Automated conversion calls the model deployment in your Microsoft Foundry resource. Its processing location depends on the deployment type, not just the resource's region:
- Global deployments can process requests across Azure regions.
- Data Zone deployments process requests within the specified data zone.
- Standard and regional provisioned deployments follow the processing-geography rules for those types.
Check the deployment's actual type and Foundry deployment-type guidance against your residency requirements. Data-at-rest commitments and inference-processing location are different considerations. Copilot requests don't inherit your Foundry deployment's residency configuration.
What do I need before I run a conversion?
On the Azure side:
- An Azure Database for PostgreSQL flexible server or Azure HorizonDB cluster that meets the tutorial prerequisites.
- An empty schema validation database or a dedicated, unused database that contains nothing you need to preserve. The tool drops schemas that correspond to the selected Oracle schemas by using
DROP SCHEMA ... CASCADEand then recreates them. This action can also remove dependent objects in other schemas. - A login with permission to connect, drop and recreate the matching schemas, and create validation objects. Arrange extension installation with your database administrator. See Required PostgreSQL privileges.
- The required PostgreSQL extensions supported by your selected service, allowed and installed in the database, and
search_pathconfigured. See the service-specific guidance in Best practices for schema conversion. - A Microsoft Foundry deployment with sufficient token-per-minute capacity. See Microsoft Foundry capacity recommendations.
On the Oracle side, see the Oracle source requirements in the tutorial.
What Oracle features can't be converted?
See Oracle to Azure PostgreSQL schema conversion limitations for known extraction gaps and partial conversions. Use the assisted correction workflow for objects with available definitions, and plan separately for objects excluded from extraction.
What does the conversion rate mean for my project?
The conversion rate is the number of objects reported as Converted divided by the conversion total, expressed as a percentage. Use the conversion totals in the HTML report, not the discovery or extraction totals.
Review and verify both converted and not-converted objects in Conversion Object Explorer to inspect their available definitions and conversion notes. The conversion rate doesn't establish equivalent runtime behavior or production readiness.
Can I rerun a completed conversion?
No. Rerunning conversion in a completed project isn't supported. Keep the generated artifacts as the reference for the project, and use Conversion Object Explorer to review and correct PostgreSQL definitions. Fix can also correct dependent objects and resolve them automatically after successful compilation. Compile manual edits and retain your corrections in postgres_ddl. Neither refreshing Discovery before conversion nor refreshing conversion reports afterward reruns a completed conversion.
What should I do if conversion fails or is interrupted?
An incomplete run is different from a completed run with objects marked Not converted.
- Preserve the schema conversion project, generated artifacts, and available logs. Record the extension version, failed stage, and error message.
- Check the project's current progress before starting another operation. Don't start a second conversion while the first is still running or recovering.
- Address the reported cause, such as an Oracle connection or client error, insufficient source permissions, lost database connectivity, expired model authentication, or throttling.
- Open the existing project and follow the actions available for its current state. Reopening the project doesn't by itself confirm that conversion resumed. If it can't continue, report the issue through PostgreSQL extension support and include redacted diagnostics.
If you choose to start a new project, preserve the original first and review the selected schema validation database again. A new conversion can drop matching schemas and their dependent objects; follow schema validation database preparation. Don't treat partial output as a complete deployment package.
For a completed run with Not converted objects, use the object correction workflow, not a conversion restart. For deployment failures, use Setup recovery guidance.
How do Resolve and Refresh differ?
Resolve in Conversion Object Explorer changes an object's recorded status to Converted. It doesn't open Copilot Chat, compile SQL, or deploy it. Refresh updates the conversion reports to reflect revised statuses. Use Refresh after manual resolution or after Fix automatically resolves successfully compiled dependent objects. These updates don't establish deployment inclusion or equivalent application behavior. See Resolve objects and refresh reports.
What validation should I do before I go to production?
Recommended validation steps:
- Review the HTML conversion report and use Conversion Object Explorer to verify both converted and not-converted objects. Read their conversion notes and resolve identified issues.
- Validate business-critical procedures, triggers, and complex views against the Oracle behavior.
- Run functional, integration, and performance tests with representative data and application workloads in the database you choose for testing.
- Verify that sequences are advanced past the data's maximum IDs before you re-enable application writes.
- Confirm that all required extensions are installed and present in
search_path.
These steps don't replace your broader migration and cutover plan. For schema-validation guidance, see Best practices for schema conversion.
How do I deploy and verify the converted schema?
Use the Setup Experience to deploy and verify the converted schema. Review the generated setup_instructions.md and save corrections in the tool-generated object files in postgres_ddl. Setup reads their latest saved contents and ignores files added manually to that directory. Prepare an Azure Database for PostgreSQL flexible server or Azure HorizonDB destination, and run validate-only checks before executing the PowerShell or Bash deployment script. Generating setup files doesn't deploy the schema.
After deployment, review the setup deployment report, object and extension verification results, and logs. Resolve any deployment errors or objects that weren't deployed. Then, smoke-test critical functions, procedures, triggers, and views in the database you choose for testing. Successful setup doesn't replace independent data and application validation.
Where can I get help?
- Documentation: Start with the schema conversion overview, tutorial, best practices, limitations, and reports articles.
- PostgreSQL extension support: File an issue on the vscode-pgsql GitHub repository and prefix the issue title with
Schema Conversion:.