Ashu Pachauri created HBASE-18125:
-------------------------------------
Summary: HBase shell disregards spaces at the end of a split key
in a split file
Key: HBASE-18125
URL: https://issues.apache.org/jira/browse/HBASE-18125
Project: HBase
Issue Type: Bug
Components: shell
Affects Versions: 1.3.1, 2.0.0
Reporter: Ashu Pachauri
When converting row keys to a printable string representation, Bytes class
considers SPACE as a printable character, so it prints it out as it is. So,
it's quite possible that a row key has a space at the end.
When specifying split points in a file, the row keys are not quoted and the
shell wrapper "admin.rb" strips any whitespace off the row keys:
{code}
File.foreach(splits_file) do |line|
arg[SPLITS].push(line.strip())
end
{code}
The correct approach is to use "chomp()" instead of "strip()" to just strip off
carriage returns and newlines. We should assume that the hbase user is either
using split points printed out by hbase itself (which will not have tabs) or is
diligent enough to not use tabs at the end of a split point.
What's worse is that it goes undetected and will result in undesirable split
points.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
