What causes Fabric Pipeline Error 4124?

Transient Azure Machine Learning service availability issues or regional outages causing HTTP errors. The referenced AzureML pipeline endpoint or resource no longer exists (e.g., deleted or unpublished). Network connectivity issues between Fabric Data Factory and the Azure Machine Learning service endpoint. AzureML service throttling due to too many concurrent requests, resulting in 429 or 503 HTTP responses

How do I fix Fabric Pipeline Error 4124?

Step 1: Check the HTTP status code and external message in the error details to narrow down the root cause — 404 suggests a missing resource, 429 or 503 suggests throttling or service unavailability, 500 indicates an AzureML-side error.. Step 2: Verify that the target Azure Machine Learning pipeline endpoint referenced in the connection still exists and is published as active in Azure Machine Learning Studio.. Step 3: Check the Azure Service Health dashboard for any ongoing incidents or degradations affecting Azure Machine Learning in your region.. Step 4: If throttling is suspected, implement retry policies on the AzureMLExecutePipeline activity in Fabric with exponential backoff to handle transient 429 errors gracefully.. Step 5: If the error is consistent and not transient, review network configurations (private endpoints, firewall rules, VNet integration) to ensure Fabric can reach the AzureML service endpoint.

High severityconnectivityMicrosoft Fabric →

Power BI Refresh Error:
Fabric Pipeline Error 4124

What does this error mean?

A request sent to Azure Machine Learning failed with an HTTP error status code during pipeline execution. This is a general AzureML communication failure that captures errors not covered by more specific error codes 4121–4123.

Common causes

1Transient Azure Machine Learning service availability issues or regional outages causing HTTP errors
2The referenced AzureML pipeline endpoint or resource no longer exists (e.g., deleted or unpublished)
3Network connectivity issues between Fabric Data Factory and the Azure Machine Learning service endpoint
4AzureML service throttling due to too many concurrent requests, resulting in 429 or 503 HTTP responses

How to fix it

1Step 1: Check the HTTP status code and external message in the error details to narrow down the root cause — 404 suggests a missing resource, 429 or 503 suggests throttling or service unavailability, 500 indicates an AzureML-side error.
2Step 2: Verify that the target Azure Machine Learning pipeline endpoint referenced in the connection still exists and is published as active in Azure Machine Learning Studio.
3Step 3: Check the Azure Service Health dashboard for any ongoing incidents or degradations affecting Azure Machine Learning in your region.
4Step 4: If throttling is suspected, implement retry policies on the AzureMLExecutePipeline activity in Fabric with exponential backoff to handle transient 429 errors gracefully.
5Step 5: If the error is consistent and not transient, review network configurations (private endpoints, firewall rules, VNet integration) to ensure Fabric can reach the AzureML service endpoint.

Frequently asked questions

Should I add retry logic to my Fabric pipeline activity to handle error 4124?

Yes, for transient errors (HTTP 429, 503, or intermittent 500s), adding retry logic with exponential backoff on the AzureMLExecutePipeline activity is strongly recommended. However, retries will not help for persistent errors like 404 (missing resource) or 403 (permissions), which require configuration fixes.

How do I tell if error 4124 is caused by an AzureML outage vs. a problem in my configuration?

Check the Azure Service Health portal for AzureML incidents in your region. If no outage is reported and the error is consistent across runs, the issue is likely configuration-related (wrong endpoint, deleted resource, or network restriction) rather than a platform outage.