关于hadoop一些自定义输出
code>OutputFormat</code> describes the output-specification for a
* Map-Reduce job.
首先继承outputFormat<key,value>这个抽象类 Map-Reduce job的输出规范
实现他的方法:
RecordWriter<KeyBaseDimension, BaseStatsValueWritable> getRecordWriter 在方法内可以进行数据库连接操作
这里需要一个返回一个RecordWriter
继承这个RecordWriter类
实现里面的write方法 进行数据库jdbc存储即可 关于reduce端输出时会调用的write方法
实现类为:TaskInputOutputContextImpl
private RecordWriter<KEYOUT,VALUEOUT> output;
public void write(KEYOUT key, VALUEOUT value
) throws IOException, InterruptedException {
output.write(key, value);
}
最终是调用了RecordWriter的write方法, map端读取hbase一个mr工具类
*在提交TableMap作业之前使用此选项。它将被适当地设置
*工作。
TableMapReduceUtil 这个类很重要,在提交读取hbase表job之前可以对其进行一系列过滤操作
FilterList filterList = new FilterList(FilterList.Operator.MUST_PASS_ALL); filterList.addFilter(
new SingleColumnValueFilter(EventLogConstants.EVENT_LOGS_FAMILY_NAME_BYTES,
Bytes.toBytes(EventLogConstants.LOG_COLUMN_NAME_EVENT_NAME),
CompareOp.EQUAL, Bytes.toBytes(EventLogConstants.EventEnum.BC_SX.alias)));
public static void initTableMapperJob(List<Scan> scans,
Class<? extends TableMapper> mapper,
Class<?> outputKeyClass,
Class<?> outputValueClass, Job job,
boolean addDependencyJars,
boolean initCredentials) throws IOException {
scan之前进行过滤器数据
List<Scan> scanList = new ArrayList<Scan>();
try {
conn = ConnectionFactory.createConnection(conf);
admin = conn.getAdmin();
String tableName = EventLogConstants.HBASE_NAME_AUDIT_SX + GlobalConstants.UNDERLINE + statDate.replaceAll(GlobalConstants.KEY_SEPARATOR, "");
if (admin.tableExists(TableName.valueOf(tableName))) {
Scan scan = new Scan(); scan读取多个表的设置
// If an application wants to use multiple scans over different tables each scan must
// define this attribute with the appropriate table name by calling
// scan.setAttribute(Scan.SCAN_ATTRIBUTES_TABLE_NAME, Bytes.toBytes(tableName))
// static public final String SCAN_ATTRIBUTES_TABLE_NAME = "scan.attributes.table.name";
scan.setAttribute(Scan.SCAN_ATTRIBUTES_TABLE_NAME, Bytes.toBytes(tableName));
scan.setFilter(filterList);
scanList.add(scan);
}
最后将job 与scanlist都设置进去 TableMapReduceUtil.initTableMapperJob(scanList, AuditorSXMapper.class,
AuditorDimensionKey.class, Text.class, job, false);
strom一些笔记知识 storm echo(File(),fun,File())
filter:实现filter接口 iskeep方法
partitionAggregate函数:分区内聚合,实现aggregate<保存聚合状态的类> 的aggregate实现聚合逻辑 ,complete方法 ridentCollector collector.emit(Value(聚合后的值))
一般的key拼接函数:实现function接口的execute方法
HBaseMapState.Options optsWait = new HBaseMapState.Options(); TridentState amtOfWaitState = partStream.project(new Fields("waitingTotalOfPartDay","dayAndContType"))
.groupBy(new Fields("dayAndContType"))
.persistentAggregate(
factoryWait,
new Fields("waitingTotalOfPartDay"),new Sum(),
new Fields("waitingGlobalOfDay")
); persistentAggregate 持久化保存函数 进行全区的sum求和,输入各区,输出为总和