Local processing job requires role, but doesn't use it #5562
Description
PySDK Version
- PySDK V2 (2.x)
- PySDK V3 (3.x)
Describe the bug
Processor.__init__ unconditionally requires an IAM role, even when running in Local Mode (instance_type="local"). The LocalSagemakerClient.create_processing_job method never uses the role (it's absorbed by **kwargs), yet the base Processor class raises a ValueError if one isn't provided. This forces users to pass a dummy role string when running processing jobs locally.
To reproduce
from sagemaker.core import FrameworkProcessor
from sagemaker.core.local import LocalSession
from sagemaker.core.helper.session_helper import Session
from sagemaker.core.image_uris import retrieve
region = Session().boto_region_name
local_session = LocalSession()
processor_image_uri = retrieve(
framework="sklearn",
version="1.4-2",
region=region
)
# This raises ValueError: "An AWS IAM role is required to create a Processing job."
processor = FrameworkProcessor(
image_uri=processor_image_uri,
instance_type="local",
instance_count=1,
sagemaker_session=local_session,
base_job_name="local-processing-test",
command=["python3"],
)Expected behavior
When instance_type is "local" or "local_gpu", the role parameter should be optional since the role is not used in Local Mode.
Screenshots or logs
╭─────────────────────────────── Traceback (most recent call last) ────────────────────────────────╮
│ in <module>:15 │
│ │
│ 12 │ region=region │
│ 13 ) │
│ 14 │
│ ❱ 15 processor = FrameworkProcessor( │
│ 16 │ image_uri=processor_image_uri, │
│ 17 │ #role="arn:aws:iam::123456789012:role/DummyRole", │
│ 18 │ instance_type="local", │
│ │
│ /opt/conda/lib/python3.12/site-packages/sagemaker/core/processing.py:1091 in __init__ │
│ │
│ 1088 │ │ if not command: │
│ 1089 │ │ │ command = ["python"] │
│ 1090 │ │ │
│ ❱ 1091 │ │ super().__init__( │
│ 1092 │ │ │ role=role, │
│ 1093 │ │ │ image_uri=image_uri, │
│ 1094 │ │ │ command=command, │
│ │
│ /opt/conda/lib/python3.12/site-packages/sagemaker/core/processing.py:755 in __init__ │
│ │
│ 752 │ │ │
│ 753 │ │ self.command = command │
│ 754 │ │ │
│ ❱ 755 │ │ super(ScriptProcessor, self).__init__( │
│ 756 │ │ │ role=role, │
│ 757 │ │ │ image_uri=image_uri, │
│ 758 │ │ │ instance_count=instance_count, │
│ │
│ /opt/conda/lib/python3.12/site-packages/sagemaker/core/processing.py:217 in __init__ │
│ │
│ 214 │ │ │ # Now we marked that as Optional because we can fetch it from SageMakerConfi │
│ 215 │ │ │ # Because of marking that parameter as optional, we should validate if it is │
│ 216 │ │ │ # after fetching the config. │
│ ❱ 217 │ │ │ raise ValueError("An AWS IAM role is required to create a Processing job.") │
│ 218 │ │ │
│ 219 │ │ self.env = resolve_value_from_config( │
│ 220 │ │ │ env, PROCESSING_JOB_ENVIRONMENT_PATH, sagemaker_session=self.sagemaker_sessi │
╰──────────────────────────────────────────────────────────────────────────────────────────────────╯
ValueError: An AWS IAM role is required to create a Processing job.System information
A description of your system. Please provide:
- SageMaker Python SDK version: 3.4.0
- Framework name (eg. PyTorch) or algorithm (eg. KMeans): scikit-learn
- Framework version: 1.4-2
- Python version: 3.10
- CPU or GPU: CPU
- Custom Docker image (Y/N): N
Additional context
The LocalSagemakerClient.create_processing_job method signature confirms the role is unused in local mode - RoleArn is not an explicit parameter and is silently discarded via **kwargs. The current workaround is to pass any non-empty string as the role (e.g., role="arn:aws:iam::123456789012:role/DummyRole"), but this is confusing for users who expect local mode to work without AWS IAM configuration.