-
Notifications
You must be signed in to change notification settings - Fork 1.2k
Closed
Labels
Description
Describe the bug
Sagemaker session always try to create default bucket, even though I already given an existing customised bucket when init Sagemaker session.
“errorMessage”: “Your previous request to create the named bucket succeeded and you already own it.“
Also, it happens continuously every few seconds.
To reproduce
import boto3
import sagemaker
from sagemaker.estimator import Estimator
bucket = "bucket-that-exists"
boto_session = boto3.session.Session(region_name="ap-southeast-1")
ml_estimator = Estimator(
image_uri = container,
role = role,
sagemaker_session = sagemaker.session.Session(boto_session, default_bucket=bucket),
**config["train_model"]["estimator_config"]
)
Expected behavior
Sagemaker should check the provided bucket is already exist and stop creating new.
However, bucket.creation_date
returns None even when the bucket exists, and trying to create it again.
Screenshots or logs
Events from cloudtrail:
“eventSource”: “s3.amazonaws.com”,
“eventName”: “CreateBucket”,
“awsRegion”: “ap-southeast-1",
“sourceIPAddress”: “3.0.220.101",
“userAgent”: “[Boto3/1.13.9 Python/3.8.5 Linux/4.14.198-152.320.amzn2.x86_64 exec-env/AWS_ECS_FARGATE Botocore/1.17.44 Resource]“,
“errorCode”: “BucketAlreadyOwnedByYou”,
“errorMessage”: “Your previous request to create the named bucket succeeded and you already own it.“,
“requestParameters”: {
“CreateBucketConfiguration”: {
“LocationConstraint”: “ap-southeast-1",
“xmlns”: “http://s3.amazonaws.com/doc/2006-03-01/”
},
“bucketName”: "bucket-that-exists",
“Host”: xxx
},
System information
- SageMaker Python SDK version: 2.10.0
- Python version: 3.8
- CPU or GPU: CPU
- Custom Docker image (Y/N): Y