com.amazonaws.services.s3.model.AmazonS3Exception: The bucket is in this region: *** when using S3 Select

[lbourgeois]

Hello,

I have a cluster running in us-east-1 region.

I hava a Spark job loading data in a DataFrame using s3select format on a bucket in eu-west-1 region.

Access and Secret keys are encoded in URI s3a://$AccessKey:$SecretKey@bucket/path/to/dir

Job fails with followong stacktrace

Caused by: com.amazonaws.services.s3.model.AmazonS3Exception: The bucket is in this region: eu-west-1. Please use this region to retry the request (Service: Amazon S3; Status Code: 301; Error Code: PermanentRedirect; Request ID: 1TTFZ54B0757A901; S3 Extended Request ID: TMqeVLFYG/b1mLVoLlSRqCMYuNbYj+cSSKneAde2/Lis7WSBvSuq98KsTcdc6SGvZHwET8GOnRs=; Proxy: null), S3 Extended Request ID: TMqeVLFYG/b1mLVoLlSRqCMYuNbYj+cSSKneAde2/Lis7WSBvSuq98KsTcdc6SGvZHwET8GOnRs=
	at com.amazonaws.http.AmazonHttpClient$RequestExecutor.handleErrorResponse(AmazonHttpClient.java:1862)
	at com.amazonaws.http.AmazonHttpClient$RequestExecutor.handleServiceErrorResponse(AmazonHttpClient.java:1415)
	at com.amazonaws.http.AmazonHttpClient$RequestExecutor.executeOneRequest(AmazonHttpClient.java:1384)
	at com.amazonaws.http.AmazonHttpClient$RequestExecutor.executeHelper(AmazonHttpClient.java:1154)
	at com.amazonaws.http.AmazonHttpClient$RequestExecutor.doExecute(AmazonHttpClient.java:811)
	at com.amazonaws.http.AmazonHttpClient$RequestExecutor.executeWithTimer(AmazonHttpClient.java:779)
	at com.amazonaws.http.AmazonHttpClient$RequestExecutor.execute(AmazonHttpClient.java:753)
	at com.amazonaws.http.AmazonHttpClient$RequestExecutor.access$500(AmazonHttpClient.java:713)
	at com.amazonaws.http.AmazonHttpClient$RequestExecutionBuilderImpl.execute(AmazonHttpClient.java:695)
	at com.amazonaws.http.AmazonHttpClient.execute(AmazonHttpClient.java:559)
	at com.amazonaws.http.AmazonHttpClient.execute(AmazonHttpClient.java:539)
	at com.amazonaws.services.s3.AmazonS3Client.invoke(AmazonS3Client.java:5453)
	at com.amazonaws.services.s3.AmazonS3Client.invoke(AmazonS3Client.java:5400)
	at com.amazonaws.services.s3.AmazonS3Client.selectObjectContent(AmazonS3Client.java:3221)
	at com.databricks.io.s3select.S3SelectDataSource$.readFileFromS3(S3SelectDataSource.scala:238)
	at com.databricks.io.s3select.S3SelectDataSource$.readFile(S3SelectDataSource.scala:284)
	at com.databricks.io.s3select.S3SelectFileFormat.$anonfun$buildReader$2(S3SelectFileFormat.scala:88)
	at org.apache.spark.sql.execution.datasources.FileFormat$$anon$1.apply(FileFormat.scala:157)
	at org.apache.spark.sql.execution.datasources.FileFormat$$anon$1.apply(FileFormat.scala:144)
	at org.apache.spark.sql.execution.datasources.FileScanRDD$$anon$1$$anon$2.getNext(FileScanRDD.scala:525)
	... 37 more

I tried to set spark.hadoop.fs.s3a.bucket.<my-bucket>.endpoint to s3.eu-west-1.amazonaws.com in cluster config without success.

Any advice ?

[Hubert-Dudek]

Maybe these resources will help:

Access S3 buckets with URIs and AWS keys https://docs.databricks.com/external-data/amazon-s3.html#access-s3-buckets-with-uris-and-aws-keys

If you are using the unity catalog and S3 buckets are inside the same account you can register them as external locations https://docs.databricks.com/data-governance/unity-catalog/manage-external-locations-and-credentials....

[lbourgeois]

Thanks @Hubert Dudek​ for having a look.

I don't use unity catalog, I actually use 3.Encode keys in URI option for S3 Auth as described in https://docs.databricks.com/external-data/amazon-s3-select.html#s3-authentication

Strange thing is that if I change format to csv in DataFrameReader I don't face this issue (even without specifying any region or endpoint). What I wonder is :

• is there any limitation around region when using s3 select connector ?
• if know how to specify a different region to avoid this exception ?

[Hubert-Dudek]

Maybe share your code as I haven't noticed s3select format and even don't know what is it 🙂

[lbourgeois]

Sure, I reproduced the issue on a notebook. Here is the code snippet to create a Dataset with s3select and csv formats :

val s3selectDS = spark.read.format("s3select").schema(mySchema)
.load("s3://"+accessKey+":"+secretKey+"@lbourgeois-rd/s3selectdbrcsv")
val csvDS = spark.read.format("csv").schema(mySchema)
.load("s3://"+accessKey+":"+secretKey+"@lbourgeois-rd/s3selectdbrcsv")

As you can see only the format arg is different.

Displaying csvDS works fine

Displaying s3selectDS raises the issue

[Hubert-Dudek]

You can try to set S3 Gateway so that it will be in the VPC network. that feature is for free https://docs.aws.amazon.com/vpc/latest/privatelink/vpc-endpoints-s3.html

https://www.databricks.com/blog/2022/11/08/optimizing-aws-s3-access-databricks.html

[Kaniz]

Hi @Laurent BOURGEOIS​(Customer)​ , We haven’t heard from you since the last response from @Hubert Dudek​ ​and I was checking back to see if their suggestions helped you.

Or else, If you have any solution, please do share that with the community as it can be helpful to others.

Also, Please don't forget to click on the "Select As Best" button whenever the information provided helps resolve your question.

[lbourgeois]

Hi @Hubert Dudek​ and @47kappal​ ,

Sorry for the delay. As suggested I'm trying to setup a gateway endpoint for s3 following https://docs.aws.amazon.com/vpc/latest/privatelink/vpc-endpoints-s3.html

I am a bit confused by

A gateway endpoint is available only in the Region where you created it. Be sure to create your gateway endpoint in the same Region as your S3 buckets.

In my case the vpc used by the cluster (and in which the gateway will be created) is us-east-1 while s3 bucket is in eu-west-1 so above statement can't be respected (bucket and gateway won't be in same region)

I am also confused by the fact that iw works with format(csv) but not with format(s3select). I wonder about limitations with s3 select connector.

[Hubert-Dudek]

It seems that you need to create vpc in another region and peer it with your main region https://aws.amazon.com/premiumsupport/knowledge-center/vpc-endpoints-cross-region-aws-services/

s3select is a completely different connector, optimized to take only part of the file from s3 bucket so it is different library

[lbourgeois]

Hello,

I tried your suggestion by setting up the peering connection between the 2 VPC but issue remains the same.

The error message

The bucket is in this region: .... please use this region to retry the request

makes me think that the root cause is not at network level but at S3 Select Spark connector level which does not use correct regional s3 enpoint.

The connector does not seem to have such property : https://docs.databricks.com/external-data/amazon-s3-select.html doc

Then I tried to set following properties at Spark level as usually suggested in such situation without any effect :

spark.conf.set("fs.s3a.endpoint","s3.eu-west-1.amazonaws.com")

spark.conf.set("fs.s3n.endpoint","s3.eu-west-1.amazonaws.com")

spark.conf.set("fs.s3.endpoint","s3.eu-west-1.amazonaws.com")

It seems that the S3 Select connector does not forward this endpoint setting to the underlying AWS S3 SDK