[HADOOP-12709] Cut s3:// from trunk

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 3.0.0-alpha1
• Fix Version/s: 3.0.0-alpha1
• Component/s: fs/s3
• Target Version/s: 3.0.0-alpha1
• Hadoop Flags: Incompatible change
• Release Note: The s3 file system has been removed. The s3a file system should be used instead.
• Assignee: Mingliang Liu
• Reporter: Steve Loughran
• Created: 14/Jan/16 18:56
• Updated: 14/Jul/16 13:33
• Resolved: 29/Jun/16 15:05

Description

UPDATE: This JIRA ticket is to track the effort of cutting the s3:// from trunk branch. Please see the cloned JIRA ticket HADOOP-13239 for deprecating s3:// from branch-2.

The fact that s3:// was broken in Hadoop 2.7 and nobody noticed until now shows that it's not being used. while invaluable at the time, s3n and especially s3a render it obsolete except for reading existing data.

I propose

1. Mark Java source as @deprecated
2. Warn the first time in a JVM that an S3 instance is created, "deprecated -will be removed in future releases"
3. in Hadoop trunk we really cut it. Maybe have an attic project (external?) which holds it for anyone who still wants it. Or: retain the code but remove the fs.s3.impl config option, so you have to explicitly add it for use.

Issue Links

• duplicates: HADOOP-10372 Deprecate S3 implementation.
• is cloned by: HADOOP-13239 Deprecate s3:// in branch-2
• is related to: HADOOP-12696 Add Tests for S3FileSystem Contract
• relates to: HADOOP-12815 TestS3ContractRootDir#testRmEmptyRootDirNonRecursive and TestS3ContractRootDir#testRmRootRecursive fail on branch-2.
• relates to: HADOOP-12801 Suppress obsolete S3FileSystem tests.