[YARN-91] DFIP aka 'NodeManager should handle Disk-Failures In Place'

• Type: Task
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: nodemanager
• Assignee: Unassigned
• Reporter: Vinod Kumar Vavilapalli
• Created: 08/Sep/12 00:13
• Updated: 04/Mar/14 19:24

Description

Moving stuff over from the MAPREDUCE JIRA: MAPREDUCE-3121

Issue Links

• relates to: YARN-257 NM should gracefully handle a full local disk

Sub-Tasks

• NM disk failure detection only covers local dirs: Open
• Add tests for testing other NM components with disk failures: Open
• NodeManager should identify failed disks becoming good again: Closed
• DefaultContainerExecutor should pick a working directory randomly: Closed