Deployment pipeline failure for a semantic model - "Backend Error: Other AS persistency error"

[gpetrites]

I have a simple deployment pipeline between two workspaces ("development" and "production"). Each workstation has a single lake house with a few semantic models that use the lake house.

I'm trying to deploy a report/semantic model. The semantic model pulls all the data from a lake house using the SQL endpoint. I was able to deploy once, which created the model in the production, but it is pointing to the development lake house. After the initial deployment, I added a deployment rule to replace the development SQL endpoint to the production endpoint. This is something that I had done previously with other semantic models, so I know how to do this correctly. Whenever thi deployment rule is in place, I am getting the following error:

Backend Error: Other AS persistency error
Operation ID: 20e016fc-1daa-4ba3-a57a-247cd9def6aa
Correlation ID: 873ef36d-23c4-f7b0-0a0f-2e0e64f69ee2
Request ID: 7169c2d2-325d-df0f-973f-742a94f98153
Time: 6/5/2024, 11:35:48 AM
Service version: 13.0.23408.46
Client version: 2405.4.19503-train
Cluster URI: https://wabi-us-east-a-primary-redirect.analysis.windows.net/

I have verified that I can connect to the SQL endpoint using SSMS.

I have even tried setting the deployment rule to replace the SQL endpoint with the exact same value (effectively not changing the endpoint) and get the same error. The only way I've found to eliminate the error is to remove the deployment rule.

I tried deleting the semantic model from production and redeploying an consistently encounter the error after adding the deployment rule.

If I edit the sematic model in "development" to use the "production" lake house, eliminating the need for a deploymet rule, and delete the production model, I can deploy to "production"...once. All future deployments of the model fail even though there is no longer a deployment rule.

I have another semantic model in workspacee that is using the same lake houses and I am able to deploy successfully multiple times with the desired deployment rule.

[gpetrites]

The linked error is about creating a semantic model. This is different than my symptoms. I am able to create the model via the deployment pipeline. The error occurs when trying to use the deployment to update the model.

I did delete both models from the production workspace and continue to experience the same issue with the same semantic model.

Both models are accessing the same lake house...or that is the objective. But since I cannot redeploy the one model, I am unable to update the lake house reference to get them both pointed at the production lake house.

Note that my work around is to deploy the model to production using Power BI Desktop. So my work around is:

- edit the model in Power BI Desktop

- update data sources to point to production lake house

- make/test changes

- publish to development so it's available for reports to use

- change reports

- commit model and reports to git

- publish updated model to production using Power BI Desktop

- use deployment pipeline to publish updated reports

While this works, it means a different deployment proces for the model and the reports and I have no way to verify the model changes have been deployed.

Also note that I tried deploying again today and got the same error. So this isn't one of those "wait a day and the error will mysteriously go away" issuues.

[razor105]

Hi @gpetrites ! Were you able to resolve this? I'm also facing the same issue.

[gpetrites · Accepted solution]

Not exactly. I temporarily gave up on the deployment pipeline. Days later on a whim I decided to try again...and it succeeded. Other than "waiting a few days" I did nothing to try to resolve.

Seems this is one of those weird errors that somehow eventually resolves on it's own.

Hopefully yours wil resolve on it's own as well.

[razor105]

Got it thanks! Hopefully it does. 

[Amélie]

Hello, 

Did you solve your issue ? I'm also facing the same error ... 

[gpetrites]

We updated our semantic model in dev to point to the production lake house. It's a bit of a hack, but it allowed us to get past the error.

[Anonymous]

Hi @gpetrites ,

Please follow these steps:

Ensure that sensitivity labels are not disabled at the organization level. For more information, please see:

Solved: Re: Get Unexpected error when creating semantic mo... - Microsoft Fabric Community

Consider removing any remnants of previous deployments, including any hidden artifacts or metadata.

Compare the workspace setup and configuration between successful and problematic semantic models.

For example, differences in data sources, permissions, etc.

Feel free to contact me if you have any further questions.

Best Regards,
Yang
Community Support Team

If there is any post helps, then please consider Accept it as the solution  to help the other members find it more quickly.
If I misunderstand your needs or you still have problems on it, please feel free to let us know. Thanks a lot!

[filip_ai]

@Anonymous I have a similar error when trying to sync a new workspace from a new feature branch in Git. It actually creates the semantic model (other than the default), and the Git status shows "Synced" for a brief second. After that the Git status changes to Uncommitted and after 10-20 seconds, the model disappears from the item list in the workspace. Eventually, the sync fails with the error below:

Workload Error Code Dataset_Import_FailedToImportDataset

Workload Error Message Other AS persistency error

I tried changing the SQL Endpoint used in the model to the new workspace's SQL Endpoint, but with no difference in behaviour:

Any advice?