[
https://issues.apache.org/jira/browse/HBASE-12757?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14257934#comment-14257934
]
pangxiaoxi commented on HBASE-12757:
------------------------------------
protected void setup(Context context){
TableSplit split = (TableSplit)context.getInputSplit();
if(split != null){
start_row = split.getStartRow();
end_row = split.getEndRow();
.....
}
}
public void map(ImmutableBytesWritable row,Result columns ,Context context) {
...
byte[] rowkey = row.get();
if(Bytes.compareTo(start_row, rowkey) > 0 || Bytes.compareTo(end_row,
rowkey) < 0){
TableSplit split = (TableSplit)context.getInputSplit();
if(split != null){
System.err.println(String.format("location=%1$s,start_row=%2$s
,rowkey=%3$s ,end_row=%4$s",
split.getRegionLocation() ,HbaseUtil.printRowkey(split.getStartRow()),
HbaseUtil.printRowkey(rowkey),HbaseUtil.printRowkey(split.getEndRow())));
}
}
.................
> MR map's input rowkey out of range of current Region
> -----------------------------------------------------
>
> Key: HBASE-12757
> URL: https://issues.apache.org/jira/browse/HBASE-12757
> Project: HBase
> Issue Type: Bug
> Components: Client, hbase
> Affects Versions: 0.94.7
> Environment: hadoop 1.1.2, r1440782
> hbase 0.94.7
> linux 2.6.32-279.el6.x86_64
> Reporter: pangxiaoxi
> Priority: Critical
>
> I excute mapreduce scan all table, sometimes map input value of rowkey is out
> of range on current Region (get from inputsplit ).
> this mabey lost data or get unused data.
> eg.
> location=datanode11,start_row=D9CB114FD09A82A3_0000000000000000_m_43DAAA689D4AFC86
> ,rowkey=D323E1D0A51E5185_0000000000000000_m_75686B8924108044
> ,end_row=DB0C4FC44E6D80C1_0000000000000000_m_E956CC65322BA3E5
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
