Unbale to run synapse pipeline when triggered from ADF, but runs inside synapse just fine.

Question

Unbale to run synapse pipeline when triggered from ADF, but runs inside synapse just fine.

Indu Jaishwal 0 Microsoft Employee

We have two Synapse pipelines inside our synapse resource which runs fine when triggered from synapse, but when triggered from ADF, they fail with the below error: Error details

This application failed due to the total number of errors: 1. Error code 1 Spark_Ambiguous_UserApp_NullPointer Message Job failed during run time with state=[dead]. TSG:The code tried to dereference a null value. At some point the code attempted to call a method or access a property on a null value. To avoid de-referencing null add null check guards around method calls or property accesses on values that can potentially be null. 1. Check the logs for this Spark application. Inspect the logs for a clearer indication of what was de-referenced. Source Unknown

The synapse resource has the following roles enabled for the ADF resource: Synapse Artifact User, Synapse Compute Operator, Synapse Credential User.
The storage account where the synapse resource stores the jar that is triggered from the synapse pipeline has granted the Storage Blob Data Contributor role to the ADF resource

Manoj Kumar Boyini 13,850 Reputation points Microsoft External Staff Moderator

2026-04-29T00:36:30.79+00:00
Hi Indu Jaishwal

Based on the behavior described, this issue is not related to permissions but to a difference in execution context when the Synapse pipeline is triggered from ADF.

Since the pipeline runs successfully within Synapse but fails from ADF with a Spark_Ambiguous_UserApp_NullPointer error, it indicates that a required value is resolving to null at runtime when triggered externally.

The most common cause is that parameters expected by the Synapse pipeline are not being passed correctly from ADF, or default values defined in Synapse are not applied during external invocation.

I recommend checking the following:

Validate that all required parameters are explicitly mapped in the ADF “Execute Synapse Pipeline” activity

Ensure parameter names match exactly between ADF and Synapse

Do not rely on default values defined in Synapse when triggering from ADF

Review the Spark application logs to identify which variable or input is resolving to null

Since the error is a NullPointerException, the failure is occurring inside the Spark job due to missing or null input rather than a platform or permission issue.
Amira Bedhiafi 41,386 Reputation points MVP Volunteer Moderator

2026-05-04T13:45:28.91+00:00

Hello Indu !

Thank you for posting on MS Learn Q&A.

I agree this looks more like a runtime context or parameter issue than a pure permission issue.

One useful next step would be to compare the successful Synapse run input with the failed ADF-triggered run input side by side, especially the pipeline parameters, Spark activity arguments, JAR path, main class, configuration values, and any dynamic expressions.

Also check whether the Synapse pipeline depends on trigger/system values, global parameters, workspace defaults, or linked-service values that exist when run manually in Synapse Studio but are not populated when invoked from ADF. For example, values like environment name, folder/date path, batch ID, file path, or config table lookup can become null and only surface later as a Spark NullPointerException.

I would also add temporary logging before the Spark/JAR activity starts: print all resolved pipeline parameters, activity inputs, JAR arguments, and config values. That will quickly show which value is different or null in the ADF-triggered execution.

One more thing to verify is the parameter data type. ADF pipeline parameter passing can sometimes behave differently for complex values, for example arrays/objects being passed differently than expected, so it is safer to explicitly pass and validate every required value rather than relying on defaults.

1 answer

Your answer

Manoj Kumar Boyini 13,850 Reputation points Microsoft External Staff Moderator

2026-04-29T00:36:30.79+00:00

Hi Indu Jaishwal

Based on the behavior described, this issue is not related to permissions but to a difference in execution context when the Synapse pipeline is triggered from ADF.

Since the pipeline runs successfully within Synapse but fails from ADF with a Spark_Ambiguous_UserApp_NullPointer error, it indicates that a required value is resolving to null at runtime when triggered externally.

The most common cause is that parameters expected by the Synapse pipeline are not being passed correctly from ADF, or default values defined in Synapse are not applied during external invocation.

I recommend checking the following:

Validate that all required parameters are explicitly mapped in the ADF “Execute Synapse Pipeline” activity

Ensure parameter names match exactly between ADF and Synapse

Do not rely on default values defined in Synapse when triggering from ADF

Review the Spark application logs to identify which variable or input is resolving to null

Since the error is a NullPointerException, the failure is occurring inside the Spark job due to missing or null input rather than a platform or permission issue.
Amira Bedhiafi 41,386 Reputation points MVP Volunteer Moderator

2026-05-04T13:45:28.91+00:00

Hello Indu !

Thank you for posting on MS Learn Q&A.

I agree this looks more like a runtime context or parameter issue than a pure permission issue.

One useful next step would be to compare the successful Synapse run input with the failed ADF-triggered run input side by side, especially the pipeline parameters, Spark activity arguments, JAR path, main class, configuration values, and any dynamic expressions.

Also check whether the Synapse pipeline depends on trigger/system values, global parameters, workspace defaults, or linked-service values that exist when run manually in Synapse Studio but are not populated when invoked from ADF. For example, values like environment name, folder/date path, batch ID, file path, or config table lookup can become null and only surface later as a Spark NullPointerException.

I would also add temporary logging before the Spark/JAR activity starts: print all resolved pipeline parameters, activity inputs, JAR arguments, and config values. That will quickly show which value is different or null in the ADF-triggered execution.

One more thing to verify is the parameter data type. ADF pipeline parameter passing can sometimes behave differently for complex values, for example arrays/objects being passed differently than expected, so it is safer to explicitly pass and validate every required value rather than relying on defaults.

Answer 1

The failure occurs because the Azure Data Factory (ADF) trigger runs the Synapse Spark job under ADF’s managed identity, which differs from the identity used when running the pipeline directly in Synapse Studio. This causes missing runtime context (linked service / Spark config / parameter values) and results in a Spark NullPointerException.

Try below Fix Steps:

Ensure the ADF managed identity has:
- Synapse Administrator or all required roles plus explicit permissions on:
- The Spark pool
- Linked services used by the pipeline
- Access to all storage paths referenced at runtime (input, output, temp).
Verify the pipeline does not rely on Studio-only defaults (parameters, Spark config) and explicitly pass all required values when triggered from ADF.

Share via

Unbale to run synapse pipeline when triggered from ADF, but runs inside synapse just fine.

1 answer

Your answer