数据湖(十七):Flink与Iceberg整合DataStream API操作

2022-07-15 08:29:08 浏览数 (1)

Flink与Iceberg整合DataStream API操作

目前Flink支持使用DataStream API 和SQL API 方式实时读取和写入Iceberg表,建议大家使用SQL API 方式实时读取和写入Iceberg表。

Iceberg 支持的Flink版本为1.11.x版本以上,目前经过测试Iceberg版本与Flink的版本对应关系如下:

  • Flink1.11.x版本与Iceberg0.11.1版本匹配。
  • Flink1.12.x~Flink1.1.x 版本与Iceberg0.12.1版本匹配,SQL API有一些bug。
  • Flink1.14.x版本与Iceberg0.12.1版本能整合但是有一些小bug,例如实时读取Iceberg中的数据有bug。

以下Flink与Iceberg整合使用的Flink版本为1.13.5,Iceberg版本为0.12.1版本。后期使用SQL API 操作时使用的Flink版本为1.11.6,Iceberg版本为0.11.1版本。

一、DataStream API 实时写入Iceberg表

DataStream Api方式操作Iceberg方式目前仅支持Java Api。使用DataStream API 实时写入Iceberg表具体操作如下:

1、首先在Maven中导入以下依赖

代码语言:javascript复制
<properties>
  <project.build.sourceEncoding>UTF-8</project.build.sourceEncoding>
  <maven.compiler.source>1.8</maven.compiler.source>
  <maven.compiler.target>1.8</maven.compiler.target>
  <!-- flink 1.12.x -1.13.x  版本与Iceberg 0.12.1 版本兼容 ,不能与Flink 1.14 兼容-->
  <flink.version>1.13.5</flink.version>
  <!--<flink.version>1.12.1</flink.version>-->
  <!--<flink.version>1.14.2</flink.version>-->
  <!-- flink 1.11.x 与Iceberg 0.11.1 合适-->
  <!--<flink.version>1.11.6</flink.version>-->
  <hadoop.version>3.2.2</hadoop.version>
</properties>

<dependencies>
    <dependency>
        <groupId>com.alibaba.ververica</groupId>
        <artifactId>ververica-connector-iceberg</artifactId>
        <version>1.13-vvr-4.0.7</version>
    </dependency>
    <!-- Flink 操作Iceberg 需要的Iceberg依赖 -->
  <dependency>
    <groupId>org.apache.iceberg</groupId>
    <artifactId>iceberg-flink-runtime</artifactId>
    <version>0.12.1</version>
    <!--<version>0.11.1</version>-->
  </dependency>

  <!-- java 开发Flink 所需依赖 -->
  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-java</artifactId>
    <version>${flink.version}</version>
  </dependency>
  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-streaming-java_2.11</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-clients_2.11</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-streaming-scala_2.11</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <!-- Flink Kafka连接器的依赖 -->
  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-connector-kafka_2.11</artifactId>
    <version>${flink.version}</version>
  </dependency>
  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-connector-base</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <!-- 读取hdfs文件需要jar包-->
  <dependency>
    <groupId>org.apache.hadoop</groupId>
    <artifactId>hadoop-client</artifactId>
    <version>${hadoop.version}</version>
  </dependency>

  <!-- Flink SQL & Table-->
  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-table-runtime-blink_2.11</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-table</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-table-common</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-table-api-java</artifactId>
    <version>${flink.version}</version>
  </dependency>
  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-table-api-java-bridge_2.11</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-table-planner_2.11</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <dependency>
    <groupId>org.apache.flink</groupId>
    <artifactId>flink-table-planner-blink_2.11</artifactId>
    <version>${flink.version}</version>
  </dependency>

  <dependency>
    <groupId>junit</groupId>
    <artifactId>junit</artifactId>
    <version>4.11</version>
    <scope>test</scope>
   </dependency>

  <!-- log4j 和slf4j 包,如果在控制台不想看到日志,可以将下面的包注释掉-->
  <dependency>
  <groupId>org.slf4j</groupId>
  <artifactId>slf4j-log4j12</artifactId>
  <version>1.7.25</version>
  <scope>test</scope>
  </dependency>
  <dependency>
  <groupId>log4j</groupId>
  <artifactId>log4j</artifactId>
  <version>1.2.17</version>
  </dependency>
  <dependency>
  <groupId>org.slf4j</groupId>
  <artifactId>slf4j-api</artifactId>
  <version>1.7.25</version>
  </dependency>
  <dependency>
  <groupId>org.slf4j</groupId>
  <artifactId>slf4j-nop</artifactId>
  <version>1.7.25</version>
  <scope>test</scope>
  </dependency>
  <dependency>
  <groupId>org.slf4j</groupId>
  <artifactId>slf4j-simple</artifactId>
  <version>1.7.5</version>
  </dependency>
</dependencies>

2、编写代码使用DataStream API将Kafka数据写入到Iceberg表

代码语言:javascript复制
import com.google.common.collect.ImmutableMap;
import org.apache.flink.api.common.eventtime.WatermarkStrategy;
import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.api.common.serialization.SimpleStringSchema;
import org.apache.flink.connector.kafka.source.KafkaSource;
import org.apache.flink.connector.kafka.source.enumerator.initializer.OffsetsInitializer;
import org.apache.flink.streaming.api.datastream.DataStreamSource;
import org.apache.flink.streaming.api.datastream.SingleOutputStreamOperator;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.table.data.GenericRowData;
import org.apache.flink.table.data.RowData;
import org.apache.hadoop.conf.Configuration;
import org.apache.iceberg.*;
import org.apache.iceberg.catalog.Catalog;
import org.apache.iceberg.catalog.TableIdentifier;
import org.apache.iceberg.flink.TableLoader;
import org.apache.flink.table.data.StringData;
import org.apache.iceberg.flink.sink.FlinkSink;
import org.apache.iceberg.hadoop.HadoopCatalog;
import org.apache.iceberg.types.Types;
import java.util.Map;

/**
 * 使用DataStream Api 向Iceberg 表写入数据
 */
public class StreamAPIWriteIceberg {
    public static void main(String[] args) throws Exception {
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();
        //1.必须设置checkpoint ,Flink向Iceberg中写入数据时当checkpoint发生后,才会commit数据。
        env.enableCheckpointing(5000);

        //2.读取Kafka 中的topic 数据
        KafkaSource<String> source = KafkaSource.<String>builder()
                .setBootstrapServers("node1:9092,node2:9092,node3:9092")
                .setTopics("flink-iceberg-topic")
                .setGroupId("my-group-id")
                .setStartingOffsets(OffsetsInitializer.latest())
                .setValueOnlyDeserializer(new SimpleStringSchema())
                .build();
        DataStreamSource<String> kafkaSource = env.fromSource(source, WatermarkStrategy.noWatermarks(), "Kafka Source");

        //3.对数据进行处理,包装成RowData 对象,方便保存到Iceberg表中。
        SingleOutputStreamOperator<RowData> dataStream = kafkaSource.map(new MapFunction<String, RowData>() {
            @Override
            public RowData map(String s) throws Exception {
                System.out.println("s = " s);
                String[] split = s.split(",");
                GenericRowData row = new GenericRowData(4);
                row.setField(0, Integer.valueOf(split[0]));
                row.setField(1, StringData.fromString(split[1]));
                row.setField(2, Integer.valueOf(split[2]));
                row.setField(3, StringData.fromString(split[3]));
                return row;
            }
        });

        //4.创建Hadoop配置、Catalog配置和表的Schema,方便后续向路径写数据时可以找到对应的表
        Configuration hadoopConf = new Configuration();
        Catalog catalog = new HadoopCatalog(hadoopConf,"hdfs://mycluster/flink_iceberg/");

        //配置iceberg 库名和表名
        TableIdentifier name =
                TableIdentifier.of("icebergdb", "flink_iceberg_tbl");

        //创建Icebeng表Schema
        Schema schema = new Schema(
                Types.NestedField.required(1, "id", Types.IntegerType.get()),
                Types.NestedField.required(2, "nane", Types.StringType.get()),
                Types.NestedField.required(3, "age", Types.IntegerType.get()),
                Types.NestedField.required(4, "loc", Types.StringType.get()));

        //如果有分区指定对应分区,这里“loc”列为分区列,可以指定unpartitioned 方法不设置表分区
//        PartitionSpec spec = PartitionSpec.unpartitioned();
        PartitionSpec spec = PartitionSpec.builderFor(schema).identity("loc").build();

        //指定Iceberg表数据格式化为Parquet存储
        Map<String, String> props =
                ImmutableMap.of(TableProperties.DEFAULT_FILE_FORMAT, FileFormat.PARQUET.name());
        Table table = null;

        // 通过catalog判断表是否存在,不存在就创建,存在就加载
        if (!catalog.tableExists(name)) {
            table = catalog.createTable(name, schema, spec, props);
        }else {
            table = catalog.loadTable(name);
        }

        TableLoader tableLoader = TableLoader.fromHadoopTable("hdfs://mycluster/flink_iceberg/icebergdb/flink_iceberg_tbl", hadoopConf);

        //5.通过DataStream Api 向Iceberg中写入数据
        FlinkSink.forRowData(dataStream)
                //这个 .table 也可以不写,指定tableLoader 对应的路径就可以。
                .table(table)
                .tableLoader(tableLoader)
                //默认为false,追加数据。如果设置为true 就是覆盖数据
                .overwrite(false)
                .build();

        env.execute("DataStream Api Write Data To Iceberg");
    }
}

以上代码有如下几个注意点:

  • 需要设置Checkpoint,Flink向Iceberg中写入Commit数据时,只有Checkpoint成功之后才会Commit数据,否则后期在Hive中查询不到数据。
  • 读取Kafka数据后需要包装成RowData或者Row对象,才能向Iceberg表中写出数据。写出数据时默认是追加数据,如果指定overwrite就是全部覆盖数据。
  • 在向Iceberg表中写数据之前需要创建对应的Catalog、表Schema,否则写出时只指定对应的路径会报错找不到对应的Iceberg表。
  • 不建议使用DataStream API 向Iceberg中写数据,建议使用SQL API。

3、在Kafka 中创建代码中指定的“flink-iceberg-topic”并启动代码生产数据

代码语言:javascript复制
# 在Kafka 中创建 flink-iceberg-topic topic
[root@node1 bin]# ./kafka-topics.sh  --zookeeper node3:2181,node4:2181,node5:2181  --create  --topic flink-iceberg-topic  --partitions 3 --replication-factor 3

创建好以上topic之后,启动代码,然后向topic中生产以下数据:

代码语言:javascript复制
[root@node1 bin]#./kafka-console-producer.sh  --topic flink-iceberg-topic --broker-list node1:9092,node2:9092,node3:9092
1,zs,18,beijing
2,ls,19,shanghai
3,ww,20,beijing
4,ml,21,shanghai

可以看到在HDFS 对应的路径中保存了对应的数据:

4、通过Hive查看保存到Iceberg中的数据

启动Hive、Hive Metastore 在Hive中创建映射Iceberg的外表:

代码语言:javascript复制
CREATE TABLE flink_iceberg_tbl  (
  id int, 
  name string,
  age int,
  loc string
)
STORED BY 'org.apache.iceberg.mr.hive.HiveIcebergStorageHandler' 
LOCATION 'hdfs://mycluster/flink_iceberg/icebergdb/flink_iceberg_tbl' 
TBLPROPERTIES ('iceberg.catalog'='location_based_table');

注意:虽然loc是分区列,创建时忽略分区列就可以,此外映射表的路径要保持与保存Iceberg数据路径一致。

通过Hive查询对应的Iceberg表中的数据,结果如下:

代码语言:javascript复制
hive> select * from flink_iceberg_tbl;
OK
2	ls	19	shanghai
3	ww	20	beijing
1	zs	18	beijing
4	ml	21	shanghai

二、​​​​​​​DataStream API 批量/实时读取Iceberg表

DataStream API 读取Iceberg表又分为批量读取和实时读取。通过方法“streaming(true/false)”来控制。

1、批量/全量读取

代码语言:javascript复制
import org.apache.flink.api.common.functions.MapFunction;
import org.apache.flink.streaming.api.datastream.DataStream;
import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
import org.apache.flink.table.data.RowData;
import org.apache.hadoop.conf.Configuration;
import org.apache.iceberg.flink.TableLoader;
import org.apache.iceberg.flink.source.FlinkSource;

/**
 * 使用DataStream Api 批量/实时 读取Iceberg 数据
 */
public class StreamAPIReadIceberg {
    public static void main(String[] args) throws Exception {
        StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

        //1.配置TableLoader
        Configuration hadoopConf = new Configuration();
        TableLoader tableLoader = TableLoader.fromHadoopTable("hdfs://mycluster/flink_iceberg/icebergdb/flink_iceberg_tbl", hadoopConf);

        //2.从Iceberg中读取全量/增量读取数据
        DataStream<RowData> batchData = FlinkSource.forRowData().env(env)
                .tableLoader(tableLoader)
                //默认为false,整批次读取,设置为true 为流式读取
                .streaming(false)
                .build();

        batchData.map(new MapFunction<RowData, String>() {
            @Override
            public String map(RowData rowData) throws Exception {
                int id = rowData.getInt(0);
                String name = rowData.getString(1).toString();
                int age = rowData.getInt(2);
                String loc = rowData.getString(3).toString();
                return id "," name "," age "," loc;
            }
        }).print();

        env.execute("DataStream Api Read Data From Iceberg");

    }
}

结果如下:

2、实时读取

代码语言:javascript复制
//当配置 streaming参数为true时就是实时读取
DataStream<RowData> batchData = FlinkSource.forRowData().env(env)
        .tableLoader(tableLoader)
        //默认为false,整批次读取,设置为true 为流式读取
        .streaming(true)
        .build();

修改以上代码并启动,向Hive 对应的Iceberg表“flink_iceberg_tbl”中插入2条数据:

在向Hive的Iceberg表中插入数据之前需要加入以下两个包:

代码语言:javascript复制
add jar /software/hive-3.1.2/lib/iceberg-hive-runtime-0.12.1.jar;
add jar /software/hive-3.1.2/lib/libfb303-0.9.3.jar;

向Hive 中Iceberg 表插入两条数据

代码语言:javascript复制
hive> insert into flink_iceberg_tbl values (5,'s1',30,'guangzhou'),(6,'s2',31,'tianjin');

插入完成之后,可以看到Flink 控制台实时读取到对应数据

三、​​​​​​​​​​​​​​指定基于快照实时增量读取数据

以上案例我们发现Flink将表中所有数据都读取出来,我们也可以指定对应的snapshot-id 决定基于哪些数据增量读取数据。

代码语言:javascript复制
DataStream<RowData> batchData = FlinkSource.forRowData().env(env)
        .tableLoader(tableLoader)
        //基于某个快照实时增量读取数据,快照需要从元数据中获取
        .startSnapshotId(4226332606322964975L)
        //默认为false,整批次读取,设置为true 为流式读取
        .streaming(true)
        .build();

结果只读取到指定快照往后的数据,如下:

四、合并data files

Iceberg提供Api将小文件合并成大文件,可以通过Flink 批任务来执行。Flink中合并小文件与Spark中小文件合并完全一样。

代码如下:

代码语言:javascript复制
import org.apache.flink.api.java.ExecutionEnvironment;
import org.apache.hadoop.conf.Configuration;
import org.apache.iceberg.Table;
import org.apache.iceberg.actions.RewriteDataFilesActionResult;
import org.apache.iceberg.catalog.Catalog;
import org.apache.iceberg.catalog.TableIdentifier;
import org.apache.iceberg.flink.TableLoader;
import org.apache.iceberg.flink.actions.Actions;
import org.apache.iceberg.hadoop.HadoopCatalog;

/**
 *  可以通过提交Flink批量任务来合并Data Files 文件。
 */
public class RewrietDataFiles {
    public static void main(String[] args) {

        ExecutionEnvironment env = ExecutionEnvironment.getExecutionEnvironment();

        //1.配置TableLoader
        Configuration hadoopConf = new Configuration();

        //2.创建Hadoop配置、Catalog配置和表的Schema,方便后续向路径写数据时可以找到对应的表
        Catalog catalog = new HadoopCatalog(hadoopConf,"hdfs://mycluster/flink_iceberg/");

        //3.配置iceberg 库名和表名并加载表
        TableIdentifier name =
                TableIdentifier.of("icebergdb", "flink_iceberg_tbl");
        Table table = catalog.loadTable(name);

        //4..合并 data files 小文件
        RewriteDataFilesActionResult result = Actions.forTable(table)
                .rewriteDataFiles()
                //默认 512M ,可以手动通过以下指定合并文件大小,与Spark中一样。
                .targetSizeInBytes(536870912L)
                .execute();
    }
}

0 人点赞