SCIM User Provisioning with Salesforce (via External Client App) results in Time out

EelkeN 0 Reputation points
2025-11-10T14:11:38.3333333+00:00

Hi there!

We set up user provisioning between Entra and Salesforce some time ago using the pre-integrated enterprise app. This allows you to enter the credentials of your admin API user in Salesforce. This works perfectly and are really happy with it.

However, we've heard that Salesforce is changing how you can use their API. This will soon be done via an External Client App that handles incoming API traffic. This means that user provisioning from Entra to Salesforce will no longer be possible using an admin username and password.

We are currently setting up a custom enterprise application where you connect via the OAuth2 Client Credentials method, using the Tenant URL, Token Endpoint, Client ID, and Client Secret we obtained from the External Client App.

The connection is successful. The problem is when we try to test with a user via provisioning on demand. It keeps loading until an error occurs:

Status Code: Request Timeout Message: Processing the HTTP request resulted in an exception. Please see the HTTP response returned by the 'Response' property of this exception for details. Web Response: A request timed out. The request was to https://(tenant).my.salesforce.com/services/scim/v2. The timeout was 60,000.00 milliseconds.

We see in the audit logs that it does indeed find the user in Salesforce and has matched it with the Entra user. So the GET method works, but as soon as it tries to perform a PATCH or POST at the "Perform Action" step, it hangs until it times out.

I've attached our attribute mapping and provision error.Attribute mapping Entra Salesforce

Provision log error

Any idea what's going wrong here? And is Microsoft aware that this change is about to take place and that the pre-registered Salesforce application will likely no longer work?

Many thanks!
Eelke

Microsoft Security | Microsoft Entra | Microsoft Entra ID
{count} votes

Your answer

Answers can be marked as 'Accepted' by the question author and 'Recommended' by moderators, which helps users know the answer solved the author's problem.