[HDFS-4993] fsck can fail if a file is renamed or deleted

• Type: Bug
• Status: Closed
• Priority: Major
• Resolution: Fixed
• Affects Version/s: 2.1.0-beta, 0.23.9, 3.0.0-alpha1
• Fix Version/s: 0.23.10, 2.1.1-beta
• Target Version/s: 0.23.10
• Hadoop Flags: Reviewed
• Assignee: Robert Parker
• Reporter: Kihwal Lee
• Created: 15/Jul/13 15:33
• Updated: 12/May/16 18:15
• Resolved: 09/Aug/13 19:21

Description

In NamenodeFsck#check(), the getListing() and getBlockLocations() are not synchronized, so the file deletions or renames at the right moment can cause FileNotFoundException and failure of fsck.

Instead of failing, fsck should continue. Optionally it can record file system modifications it encountered, but since most modifications during fsck are not detected, there might be little value in recording these specifically.