Details
-
Sub-task
-
Status: Resolved
-
Major
-
Resolution: Duplicate
-
None
-
None
-
None
Description
If a block in a block group is reported by Datanode to Namenode, Namenode will check if it's corrupt. Currently BlockManager uses the size of block group to compare with the size of reported block, which makes the block is treated as corrupt block by Namenode. I think Namenode should first calculate the size of the block and then compare it with reported value
Attachments
Issue Links
- duplicates
-
HDFS-8120 Erasure coding: created util class to analyze striped block groups
- Resolved