[hbase] End key is incorrectly assigned in many region splits
-------------------------------------------------------------
Key: HADOOP-2451
URL: https://issues.apache.org/jira/browse/HADOOP-2451
Project: Hadoop
Issue Type: Bug
Components: contrib/hbase
Affects Versions: 0.16.0
Reporter: Jim Kellerman
Assignee: Jim Kellerman
Priority: Critical
Fix For: 0.16.0
When a region is split, the new child region for the upper half of the key
range is always assigned as null (meaning end of table). If splits only happen
at the end of the table, this is harmless. However under a random write load,
an inner region is just as likely to split.
This would explain why we have seen cases where after a split, there are
multiple regions whose end key is the end of the table (null)
--
This message is automatically generated by JIRA.
-
You can reply to this email to add a comment to the issue online.
