hbase-dev mailing list archives

Site index · List index
Message view « Date » · « Thread »
Top « Date » · « Thread »
From "Randy Hu (JIRA)" <j...@apache.org>
Subject [jira] [Created] (HBASE-15287) org.apache.hadoop.hbase.mapreduce.RowCounter returns incorrect result with binary row key inputs
Date Wed, 17 Feb 2016 22:02:19 GMT
Randy Hu created HBASE-15287:
--------------------------------

             Summary: org.apache.hadoop.hbase.mapreduce.RowCounter returns incorrect result
with binary row key inputs
                 Key: HBASE-15287
                 URL: https://issues.apache.org/jira/browse/HBASE-15287
             Project: HBase
          Issue Type: Bug
          Components: mapreduce, util
    Affects Versions: 1.1.1
            Reporter: Randy Hu


org.apache.hadoop.hbase.mapreduce.RowCounter takes optional start/end key as inputs (-range
option). It would work only when the string representation of value is identical to the string.
When row key is binary,  the string representation of the value would look like this: "\x00\x01",
which would be incorrect interpreted as 8 char string in the current implementation:

https://github.com/apache/hbase/blob/master/hbase-server/src/main/java/org/apache/hadoop/hbase/mapreduce/RowCounter.java

To fix that, we need change how the value is converted from command line inputs:

Change 
      scan.setStartRow(Bytes.toBytes(startKey));
to
      scan.setStartRow(Bytes.toBytesBinary(startKey));

Do the same conversion to end key as well.

The issue was discovered when the utility was used to calcualte row distribution on regions
from table with binary row keys. The hbase:meta contains the start key of each region in format
of above example. 




--
This message was sent by Atlassian JIRA
(v6.3.4#6332)

Mime
View raw message