We've created an application using SCIM 2 SDK from PingIdentity for provisioning with Azure AD. Custom mapping is set up and working.
However, when the user is CREATED, all of the fields are included in the import, but only a few fields are included in the provisioning step and sent to our application. Provisioning needs to run a second time on that user to UPDATE in order for all the fields to be included. Amongst other things, this means that first and last name are not split and it only sends the displayname (which ends up as firstname on our end).
For some users in normal provisioning, it can take days between the create and update runs so we're missing data for a long time.
Anyone know how can we can test for what's causing this and solve it so all the fields are included in the initial CREATE run for a user?
Here are the attribute mapping settings: https://imgur.com/ypfAAmD
And an example log of when the user is created with only basic fields: https://imgur.com/iOXACJh
vs. when the user is updated with all the other fields: https://imgur.com/UqDNyCv
I'm a product manager at Microsoft that works on the provisioning service and our SCIM client.
The behavior you're seeing occurs when you have attributes that are not part of the SCIM core schema included as "short" names. Attributes not defined in the SCIM core schema (RFC 7643) should have full URN syntax. Something to the effect of urn:ietf:params:scim:schemas:extension:appName:2.0:User:attributeName is commonly used by other implementations. The shaky behavior you're seeing where the AAD provisioning service fails to send these attribute values via a POST but later includes them in a PATCH comes down to different code paths in the AAD provisioning service, and the PATCH code happens to handle this differently than the POST code. This is purely by chance, however, and isn't an intentional design choice. At some future point I'm hoping we'll make this more consistent and disallow incorrectly structured attribute names entirely.
If you adjust your attribute names to align with the guidance in the SCIM spec's schema RFC and provide the attributes with fully defined URNs, you should see consistent behavior that works on both POST and PATCH.