Hadoop与HBase中遇到的问题(续)java.io.IOException: Non-increasing Bloom keys异常

在使用Bulkload向HBase导入数据中, 自己编写Map与使用KeyValueSortReducer生成HFile时, 出现了下面的异常:

java.io.IOException: Non-increasing Bloom keys: 201301025200000000000003520000000000000500 after 201311195100000000000000010000000000001600

at org.apache.hadoop.hbase.regionserver.StoreFile$Writer.appendGeneralBloomfilter(StoreFile.java:869)

at org.apache.hadoop.hbase.regionserver.StoreFile$Writer.append(StoreFile.java:905)

at org.apache.hadoop.hbase.mapreduce.HFileOutputFormat$1.write(HFileOutputFormat.java:180)

at org.apache.hadoop.hbase.mapreduce.HFileOutputFormat$1.write(HFileOutputFormat.java:136)

at org.apache.hadoop.mapred.ReduceTask$NewTrackingRecordWriter.write(ReduceTask.java:586)

at org.apache.hadoop.mapreduce.TaskInputOutputContext.write(TaskInputOutputContext.java:80)

at org.apache.hadoop.hbase.mapreduce.KeyValueSortReducer.reduce(KeyValueSortReducer.java:53)

at org.apache.hadoop.hbase.mapreduce.KeyValueSortReducer.reduce(KeyValueSortReducer.java:36)

at org.apache.hadoop.mapreduce.Reducer.run(Reducer.java:177)

at org.apache.hadoop.mapred.ReduceTask.runNewReducer(ReduceTask.java:649)

at org.apache.hadoop.mapred.ReduceTask.run(ReduceTask.java:418)

at org.apache.hadoop.mapred.Child$4.run(Child.java:255)

at java.security.AccessController.doPrivileged(Native Method)

at javax.security.auth.Subject.doAs(Subject.java:396)

at org.apache.hadoop.security.UserGroupInformation.doAs(UserGroupInformation.java:1190)

at org.apache.hadoop.mapred.Child.main(Child.java:249)

该异常在源码的StoreFile类中, 即在使用StoreFile类生成HFile文件时抛出异常, 根据控制台异常信息可以知道异常出现在源码StoreFile.java:905行处,此处是append方法,该方法调用appendGeneralBloomfilter方法,生成Bloom key, 源码为:

public static class HFileGenerateMapper extends
Mapper<LongWritable, Text, ImmutableBytesWritable, KeyValue> {
private static int familyIndex = 0;
private static Configuration conf = null;
private static MyMD5 md5 = new MyMD5();
@Override
protected void setup(Context context) throws IOException,
InterruptedException {
conf = context.getConfiguration();
familyIndex = conf.getInt("familyIndex",0);
}
@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
ImmutableBytesWritable mykey = new ImmutableBytesWritable(
value.toString().split(",")[0].getBytes());
List<KeyValue> list = null;
list = createKeyValue(value.toString());
Iterator<KeyValue> it = list.iterator();
while (it.hasNext()) {
KeyValue kv = new KeyValue();
kv = it.next();
if (kv != null) {
context.write(mykey, kv);
}
}

}
/**
* a.CITY_NO,to_char(DT,'yyyy-MM-dd'),DATA_TYPE,E0,E1,E2,E3,E4,E5,
* MEASUREPOINTID,TRANSFORMERID,ZONEID,CAPACITY
* @param str
* @return
*/
private List<KeyValue> createKeyValue(String str) {
List<KeyValue> list = new ArrayList<KeyValue>(CONSTANT_HBASE.TB2_FNColNames[familyIndex].length);
String[] values = str.toString().split(",");
String[] qualifiersName = CONSTANT_HBASE.TB2_FNColNames[familyIndex];
for (int i = 0; i < qualifiersName.length; i++) {
//需要作为rowKey的各个字段字符串组成RowKey
String rowkey = values[1]+values[0]+values[11]+values[12];
//加上32位的MD5
rowkey += md5.getMD5Code(rowkey);
String family = CONSTANT_HBASE.TB2_FamilyNames[familyIndex];
String qualifier = qualifiersName[i];
String value_str = values[i+CONSTANT_HBASE.TB2_FNColIndex[familyIndex]-1];

KeyValue kv = new KeyValue(Bytes.toBytes(rowkey),
Bytes.toBytes(family), Bytes.toBytes(qualifier),
CONSTANT_HBASE.timeStamp, Bytes.toBytes(value_str));
list.add(kv);
}
return list;
}
}

关键出错的那一句在

ImmutableBytesWritable rowkey = new ImmutableBytesWritable(value.toString().split(",")[0].getBytes());

因为最终导入RowKey的是由多个字段的字符串+32位的MD5值拼接而成的,但是生成ImmutableBytesWritable mykey却只用到第一个字段的字符串,而这个key是用来全局排序用的,所以需要mykey与KeyValue kv 的rowkey相等, 于是更改方法便是将map方法代码改成如下:

@Override
protected void map(LongWritable key, Text value, Context context)
throws IOException, InterruptedException {
List<KeyValue> list = null;
list = createKeyValue(value.toString());
Iterator<KeyValue> it = list.iterator();
while (it.hasNext()) {
KeyValue kv = new KeyValue();
kv = it.next();
if (kv != null) {
<span style="color:#FF0000;">context.write(new ImmutableBytesWritable(kv.getKey()), kv);</span>
}
}

}

运行之后成功了,可以通过http://localhost:50030/jobtracker.jsp查看任务运行状态.