主要记录下streaming模式下动态分区怎么写文件，sql模式直接写就是了，streaming模式需要自己写下分区方法。大致的数据流程是消费kafka，拆解json，数据写到hdfs(sequenceFile)路径。

1、分区需要自定义，这里是读取流数据，获取分区字段

package partitionassigner;

import org.apache.flink.api.java.tuple.Tuple2;
import org.apache.flink.core.io.SimpleVersionedSerializer;
import org.apache.flink.streaming.api.functions.sink.filesystem.BucketAssigner;
import org.apache.flink.streaming.api.functions.sink.filesystem.bucketassigners.SimpleVersionedStringSerializer;
import org.apache.hadoop.io.LongWritable;
import org.apache.hadoop.io.Text;

import java.text.SimpleDateFormat;
import java.util.Date;

public class PartitionAssigner<IN> implements BucketAssigner<Tuple2<LongWritable, Text>, String> {

    @Override
    public String getBucketId(Tuple2<LongWritable, Text> textTuple2, Context context) {
        String allCol = textTuple2.getField(1).toString();
        //截出来分区字段
        String[] strings = allCol.split("\\^");
        //取出来时间戳字段
        Long createTime = Long.parseLong(strings[10]);
        //取出来rank分区字段
        int rankPt =  Integer.parseInt(strings[11]);
        //时间戳 -> yyyyMMdd
        Date date = new Date(createTime);
        SimpleDateFormat simpleDateFormat = new SimpleDateFormat ("yyyyMMdd");
        return "day=" + simpleDateFormat.format(date) + "/" + "rank_pt=" + rankPt;
    }

    @Override
    public SimpleVersionedSerializer<String> getSerializer() {
        return SimpleVersionedStringSerializer.INSTANCE;
    }


}

2、文件输出调用分区生成方法

        StreamingFileSink<Tuple2<LongWritable, Text>> sink = StreamingFileSink.forBulkFormat(
                        path,
                        new SequenceFileWriterFactory<>(hadoopConf, LongWritable.class, Text.class,"org.apache.hadoop.io.compress.GzipCodec",
                                SequenceFile.CompressionType.BLOCK))
                .withBucketAssigner(new PartitionAssigner())
                .build();

3、我的路径写的本地，可以看到文件夹，生产上改下文件路径就好

总结：这个处理是想在流上把数据处理好写到离线数仓的dwd层，省去离线的处理逻辑，流批真正一体要做的工作还是比较多啊，维护成本当下看也高。