Azure AI Search can't connect to Azure SQL Database

Question

Azure AI Search can't connect to Azure SQL Database

George R 40

Hi there,

I created a Search service (Foundry IQ) service called "pim-ai-resource" and can't connect my SQL data to it. even though my IP is allowed and the User ID is correct. The Networking settings of both allow my IP and both have "Allow Azure Services to connect" ticked. See attached for the picture of the error.

I want to connect a View (vw_ProductSearchIndex), but with the error I tried the database connection and it still failed. User's image

Aditya N 3,890 Reputation points Microsoft External Staff Moderator

2026-07-02T15:52:37.52+00:00
Hello @George R

Thank you for reaching out Microsoft Q&A. We're looking into your issue that an Azure AI Search data source connection to an Azure SQL Database is failing when you try to connect (even after allowing your IP, using the correct user, and ticking “Allow Azure services to connect” on both sides). The screenshot shows a generic “failed to fetch”/stack trace style error, which is common when the underlying issue is network/connectivity to the SQL endpoint rather than (only) credentials.

Official Recommended Troubleshooting Steps

1. Configure Azure SQL Firewall:

On your Azure SQL Server > Networking tab:

Enable Public network access.

Check Allow Azure services and resources to access this server.

Add your client IP address under firewall rules.

Microsoft notes that even with "Allow Azure services" enabled, the portal's metadata fetch can still fail if specific rules are too restrictive.

Reference:
https://learn.microsoft.com/en-us/azure/search/search-how-to-index-sql-database

Kindly let us know if the above helps or you need further assistance on this issue.

Please "upvote" if the information helped you. This will help us and others in the community as well.
George R 40 Reputation points

2026-07-06T08:32:58.62+00:00

Hi Aditya,

I configured the Azure SQL Firewall as per your instructions and have the same error
Aditya N 3,890 Reputation points Microsoft External Staff Moderator

2026-07-18T22:57:55.69+00:00

Hello @George R

Please could you retry once. Could be browser issue too. Disable extensions(if any) and try once.

Is it possible to track changes using "High Water Mark Change Detection policy" ? As you selected View and SQL Integrated Change Tracking is not supported on views in Azure SQL.

Could you please let us know if SQL server public-access enabled or private-endpoint only?

Are you using a connection string or a managed identity?

Does AI Search and SQL in the same region/subscription?
Aditya N 3,890 Reputation points Microsoft External Staff Moderator

2026-07-22T00:51:39.93+00:00

Hello @George R

Please could you let us know if you had a chance to review the suggestion provided.
George R 40 Reputation points

2026-07-23T08:23:49.96+00:00

I manually found a another way to solve it though the original problem wasn't solved.

A Summary: The Azure portal's "Import and vectorize data" wizard kept failing — first a browser-side "Failed to fetch" error, then it couldn't see my Azure OpenAI service in the dropdown (an enumeration bug)

To get around it, I abandoned the wizard and built the four search objects (data source, index, skillset, indexer) directly via a PowerShell script, It used the search service's admin key — a resource-level credential — which authenticates against the service rather than the directory, so it sidestepped the permission bug entirely. The script ran cleanly and indexed all my products.

Answer accepted by question author

Praneeth Maddali 12,670 Microsoft External Staff Moderator

Hi @George R

As discussed with you offline on Teams, we understand that the issue was related to the Azure portal's Import and Vectorize Data wizard. The wizard initially failed with a "Failed to fetch" error and was also unable to discover the Azure OpenAI resource in the dropdown, which appears to be related to a portal-side enumeration/permission issue.

As a workaround, you successfully bypassed the portal wizard and created the required Azure AI Search components (Data Source, Index, Skillset, and Indexer) using a PowerShell script and REST APIs. The script authenticated using the Azure AI Search Admin Key, which uses resource-level authentication and avoided the portal permission issue. The script completed successfully, and all product data was indexed as expected.

Since the workaround has resolved the issue and the desired functionality is now working, we will proceed with case closure. If you need any further assistance in the future, please feel free to reach out.

Thank you for your time and cooperation.

Please do click "Accept the answer” and Yes, this can be beneficial to other community members.

If you have any other questions, let me know in the "comments" and I would be happy to help you

0 comments

1 additional answer

Your answer

Aditya N 3,890 Reputation points Microsoft External Staff Moderator

2026-07-02T15:52:37.52+00:00

Hello @George R

Thank you for reaching out Microsoft Q&A. We're looking into your issue that an Azure AI Search data source connection to an Azure SQL Database is failing when you try to connect (even after allowing your IP, using the correct user, and ticking “Allow Azure services to connect” on both sides). The screenshot shows a generic “failed to fetch”/stack trace style error, which is common when the underlying issue is network/connectivity to the SQL endpoint rather than (only) credentials.

Official Recommended Troubleshooting Steps

1. Configure Azure SQL Firewall:

On your Azure SQL Server > Networking tab:

Enable Public network access.

Check Allow Azure services and resources to access this server.

Add your client IP address under firewall rules.

Microsoft notes that even with "Allow Azure services" enabled, the portal's metadata fetch can still fail if specific rules are too restrictive.

Reference:
https://learn.microsoft.com/en-us/azure/search/search-how-to-index-sql-database

Kindly let us know if the above helps or you need further assistance on this issue.

Please "upvote" if the information helped you. This will help us and others in the community as well.
George R 40 Reputation points

2026-07-06T08:32:58.62+00:00

Hi Aditya,

I configured the Azure SQL Firewall as per your instructions and have the same error
Aditya N 3,890 Reputation points Microsoft External Staff Moderator

2026-07-18T22:57:55.69+00:00

Hello @George R

Please could you retry once. Could be browser issue too. Disable extensions(if any) and try once.

Is it possible to track changes using "High Water Mark Change Detection policy" ? As you selected View and SQL Integrated Change Tracking is not supported on views in Azure SQL.

Could you please let us know if SQL server public-access enabled or private-endpoint only?

Are you using a connection string or a managed identity?

Does AI Search and SQL in the same region/subscription?
Aditya N 3,890 Reputation points Microsoft External Staff Moderator

2026-07-22T00:51:39.93+00:00

Hello @George R

Please could you let us know if you had a chance to review the suggestion provided.
George R 40 Reputation points

2026-07-23T08:23:49.96+00:00

I manually found a another way to solve it though the original problem wasn't solved.

A Summary: The Azure portal's "Import and vectorize data" wizard kept failing — first a browser-side "Failed to fetch" error, then it couldn't see my Azure OpenAI service in the dropdown (an enumeration bug)

To get around it, I abandoned the wizard and built the four search objects (data source, index, skillset, indexer) directly via a PowerShell script, It used the search service's admin key — a resource-level credential — which authenticates against the service rather than the directory, so it sidestepped the permission bug entirely. The script ran cleanly and indexed all my products.

Answer 1

Christos Panagiotidis 3,551

The screenshot shows an unsupported combination: vw_ProductSearchIndex is a view, but SQL Integrated Change Tracking Policy is selected. That policy works only with tables.

For a view, select High Water Mark Change Detection and expose a monotonically increasing column in the view (preferably rowversion, otherwise a reliably updated timestamp). If you need SQL integrated change tracking and automatic deletion detection, index a table instead.

Also confirm that agent_readonly is a contained user in DocumentIntelligenceDB with SELECT on the view, and that the view returns a single unique, non-null column suitable as the search index key. If the Foundry wizard still shows the generic browser-side Failed to fetch, create the SQL data source under the Search service's Data sources page and run an indexer; its execution history will show the actual SQL or network error.

Christos Panagiotidis 3,551 Reputation points

2026-07-15T08:18:03.1766667+00:00

The screenshot shows an unsupported combination: vw_ProductSearchIndex is a view, but SQL Integrated Change Tracking Policy is selected. That policy works only with tables.

For a view, select High Water Mark Change Detection and expose a monotonically increasing column in the view (preferably rowversion, otherwise a reliably updated timestamp). If you need SQL integrated change tracking and automatic deletion detection, index a table instead.

Also confirm that agent_readonly is a contained user in DocumentIntelligenceDB with SELECT on the view, and that the view returns a single unique, non-null column suitable as the search index key. If the Foundry wizard still shows the generic browser-side Failed to fetch, create the SQL data source under the Search service's Data sources page and run an indexer; its execution history will show the actual SQL or network error.