Flink1.8批量Sink到HBase

实现背景：
  消费Kafka数据写入HBase时，单条处理效率太低。需要批量插入hbase,这里自定义时间窗口countWindowAll 实现100条hbase插入一次Hbase

前面我就不写了 直接上核心代码吧

/*每10秒一个处理窗口*/
DataStream> putList = filterData.countWindowAll(Constants.windowCount).apply(new AllWindowFunction, GlobalWindow>() {
    @Override
    public void apply(GlobalWindow window, Iterable message, Collector> out) throws Exception {
        List putList=new ArrayList();
        for (String value : message)
        {
            String rowKey=value.replace("::","_");
            Put put = new Put(Bytes.toBytes(rowKey.toString()));
            String[] column=value.split("::");
            for (int i = 0; i < column.length; i++) {
                put.addColumn(Bytes.toBytes(Constants.columnFamily),Bytes.toBytes(Constants.columnArray[i]),Bytes.toBytes(column[i]));
            }
            putList.add(put);
        }
        out.collect(putList);
    }
}).setParallelism(4);

putList.addSink(new HBaseSinkFunction()).setParallelism(1);

写HBase自定义Sink为

HBaseSinkFunction extends RichSinkFunction>{
@Override
public void open(Configuration parameters) throws Exception {
    super.open(parameters);
    HbaseUtils.connectHbase();
    TableName table=TableName.valueOf(Constants.tableNameStr);
    Admin admin = HbaseUtils.connection.getAdmin();
    if(!admin.tableExists(table)){
        HTableDescriptor tableDescriptor = new HTableDescriptor(Constants.tableNameStr);
        tableDescriptor.addFamily(new HColumnDescriptor(Constants.columnFamily));
        admin.createTable(tableDescriptor);
    }
}
@Override
public void invoke(List putList, Context context) throws Exception {
    Table table=HbaseUtils.connection.getTable(TableName.valueOf(Constants.tableNameStr));
    table.put(putList);
}
@Override
public void close() throws Exception {
    super.close();
    HbaseUtils.closeHBaseConnect();
}
}

看完qing