[HADOOP-19044] AWS SDK V2 - Update S3A region logic

• Type: Sub-task
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.4.0
• Fix Version/s: 3.4.0, 3.5.0, 3.4.1
• Component/s: fs/s3
• Labels: pull-request-available
• Assignee: Viraj Jasani
• Reporter: Ahmar Suhail
• Created: 18/Jan/24 11:10
• Updated: 11/Feb/24 01:09
• Resolved: 02/Feb/24 17:07

Description

If both fs.s3a.endpoint & fs.s3a.endpoint.region are empty, Spark will set fs.s3a.endpoint to 

s3.amazonaws.com here:

https://github.com/apache/spark/blob/9a2f39318e3af8b3817dc5e4baf52e548d82063c/core/src/main/scala/org/apache/spark/deploy/SparkHadoopUtil.scala#L540 

 

HADOOP-18908, updated the region logic such that if fs.s3a.endpoint.region is set, or if a region can be parsed from fs.s3a.endpoint (which will happen in this case, region will be US_EAST_1), cross region access is not enabled. This will cause 400 errors if the bucket is not in US_EAST_1. 

 

Proposed: Updated the logic so that if the endpoint is the global s3.amazonaws.com , cross region access is enabled.  

 

 

Issue Links

• causes: HADOOP-19066 AWS SDK V2 - Enabling FIPS should be allowed with central endpoint
• is related to: SPARK-46793 Revert S3A endpoint fixup logic of SPARK-35878
• links to: GitHub Pull Request #6479
• links to: GitHub Pull Request #6482
• links to: GitHub Pull Request #6524