[HADOOP-7659] fs -getmerge isn't guaranteed to work well over non-HDFS filesystems

• Type: Improvement
• Status: Resolved
• Priority: Minor
• Resolution: Fixed
• Affects Version/s: 0.20.204.0
• Fix Version/s: 3.0.0-alpha1
• Component/s: fs
• Release Note: Documented that the "fs -getmerge" shell command may not work properly over non HDFS-filesystem implementations due to platform-varying file list ordering.
• Assignee: Harsh J
• Reporter: Harsh J
• Created: 20/Sep/11 07:05
• Updated: 12/May/16 18:21
• Resolved: 25/May/12 12:48

Description

When you use fs -getmerge with HDFS, you are guaranteed file list sorting (part-00000, part-00001, onwards). When you use the same with other FSes we bundle, the ordering of listing is not guaranteed at all. This is cause of http://download.oracle.com/javase/6/docs/api/java/io/File.html#list() which we use internally for native file listing.

This should either be documented as a known issue on -getmerge help pages/mans, or a consistent ordering (similar to HDFS) must be applied atop the listing. I suspect the latter only makes it worthy for what we include - while other FSes out there still have to deal with this issue. Perhaps we need a recommendation doc note added to our API?