Forum Discussion
InvalidRequest when adding "Private Link Service" as a "Managed Private Endpoint"
- 2 years ago
This is extremely frustrating. Here is the support response:
Hello Josh,
I hope you are having an amazing day.
We have shared the issue with all collected details in our SME channel. They responded that currently, private endpoints targeting Private Link Service are not supported. Additionally, we do not support creating managed private endpoints with FQDNs at this time and there is no ETA. The product team will add this limitation to public documents once they have an ETA.
I am pleased to inform you that there is no issue from your side; however, the actual feature is not supported currently. We have already created an incident with the product team, and they are working on updating the feature.
I apologize for any inconvenience this may cause. Rest assured, you will be able to access this feature soon.
Please feel free to reach out to me if you have any questions, concerns, or additional information that you would like to share. I'll be more than happy to assist you.
Thank you for your understanding.
Digidank
Do you have a ticket open? DId they update the docs to say that PLS is not supported?
The docs are pretty poor. The only relevant thing I can find is like so:
Creating a managed private endpoint with a fully qualified domain name (FQDN) is not supported.
These limitations and considerations might affect your use cases and workflows. Take them into account
This doesn't specifically talk about PLS, but the FQDN part is often a show-stopper for PLS.
Not sure if anyone else has tried to create a PLS private endpoint by going around the back door. But it doesn't work, from what I've tried. Image:
I've been able to create PE requests for blob and other resources, but it never succeeds when I try to create one for PLS, with a blank group ID and a FQDN. I think we are at the mercy of whatever team is sitting on this request.
IMO, It seems like a pretty basic requirement. Private link service is a pretty fundamental component in azure nowadays, and was intended to be used for precisely this sort of scenario.
Again, it took the Synapse team about a year to create the U/I that allows this PLS connectivity to work, and I am guessing the Fabric team will take the same a similar amount of time. It would be nice if they could give a "back-door" approach that circumvents the U/I, as a temporary solution. For some reason U/I teams in the cloud seem to be extremely slow at what they do. Back-door options include a REST api, or powershell command or similar.
- dbeavon31 year ago
Memorable Member
Wahted to re-iterate that a pipeline activity "copy activity" might be the best way to work around the missing PLS. Ie. you have to go thru a round-about approach that sends the data from a service thru an ADLS storage account (temp files) before it is accessible from Fabric.
This unfortunately can cost more money, and has more moving parts (on premise gateway).
More details on reddit:
https://www.reddit.com/r/MicrosoftFabric/comments/1er2z51/invalidrequest_when_adding_a_private_link_service/ - dbeavon31 year ago
Memorable Member
I was able to get in touch with a Mindtree engineer about this today (12/17/2024):
I have an update for you from my PG, they informed me that as of now there is no ETA for this deployment as currently there is a deployment freeze. And they have provided below platforms where we can check for the latest update directly from the product team.
And I also checked for workaround but they informed me that there is not workaround and PG team is working on it to bring it as soon as possible.
https://blog.fabric.microsoft.com/en-us/blog/
https://blog.fabric.microsoft.com/en-us/blog/category/roadmap
As a side, I looked at the roadmap and it goes to the end of Q1 2025. It is reasonable to assume they are NOT committed to fixing the PLS private endpoints before the end of Q1. Unfortunately we are already planning to get workloads out of Synapse before then... so we must stumble our way thru this in one way or another. I'm guessing we will have to retrieve this service-hosted data by way of data-factory-pipelines or some other wierdness. I guess this is par for the course, when building a solution inside of a SaaS.