[MAPREDUCE-2436] RAID block fixer should prioritize block fix operations

• Type: Improvement
• Status: Resolved
• Priority: Major
• Resolution: Won't Fix
• Affects Version/s: 0.20.2, 0.20.3
• Component/s: contrib/raid
• Assignee: Ramkumar Vadali
• Reporter: Ramkumar Vadali
• Created: 13/Apr/11 19:15
• Updated: 01/Aug/17 17:12
• Resolved: 01/Aug/17 17:12

Description

The RAID block fixer submits mapreduce jobs to fix corrupt files. This is OK for XOR RAID, but with Reed-Solomon RAID, there can be large number of corrupt files when even a single datanode goes dead. With Reed-SOlomon RAID, it is better to categorize corrupt files based on urgency. Files with only one corrupt block can be treated as lower priority than those with more number of corrupt blocks.