将 Apache Hudi 与 Apache Flink 结合使用 - Amazon EMR
提交 Apache Hudi 作业

将 Apache Hudi 与 Apache Flink 结合使用

Apache Hudi 是一个开源数据管理框架，包含插入、更新、更新插入和删除等记录级操作，可用于简化数据管理和数据管道开发。与 Amazon S3 中的高效数据管理相结合，Hudi 允许实时摄取和更新数据。Hudi 会维护在数据集上运行的所有操作的元数据，因此所有操作都能保持原子性和一致性。

Apache Hudi 已在 Amazon EMR on EKS 上投入使用，搭配 Apache Flink 和 Amazon EMR 7.2.0 及更高版本。请参阅以下步骤，了解如何开始和提交 Apache Hudi 作业。

请参阅以下步骤，了解如何提交 Apache Hudi 作业。

  1. 创建一个名为 default 的 AWS Glue 数据库。

    aws glue create-database --database-input "{\"Name\":\"default\"}"

  2. 按照 Flink Kubernetes Operator SQL 示例构建 flink-sql-runner.jar 文件。

  3. 创建如下所示的 Hudi SQL 脚本。

    CREATE CATALOG hudi_glue_catalog WITH (
'type' = 'hudi',
'mode' = 'hms',
'table.external' = 'true',
'default-database' = 'default',
'hive.conf.dir' = '/glue/confs/hive/conf/',
'catalog.path' = 's3://<hudi-example-bucket>/FLINK_HUDI/warehouse/'
);

USE CATALOG hudi_glue_catalog;
CREATE DATABASE IF NOT EXISTS hudi_db;
use hudi_db;

CREATE TABLE IF NOT EXISTS hudi-flink-example-table(
    uuid VARCHAR(20),
    name VARCHAR(10),
    age INT,
    ts TIMESTAMP(3),
    `partition` VARCHAR(20)
)
PARTITIONED BY (`partition`)
WITH (
  'connector' = 'hudi',
  'path' = 's3://<hudi-example-bucket>/hudi-flink-example-table',
  'hive_sync.enable' = 'true',
  'hive_sync.mode' = 'glue',
  'hive_sync.table' = 'hudi-flink-example-table',
  'hive_sync.db' = 'hudi_db',
  'compaction.delta_commits' = '1',
  'hive_sync.partition_fields' = 'partition',
  'hive_sync.partition_extractor_class' = 'org.apache.hudi.hive.MultiPartKeysValueExtractor',
  'table.type' = 'COPY_ON_WRITE'
);

EXECUTE STATEMENT SET
BEGIN

INSERT INTO hudi-flink-example-table VALUES
    ('id1','Alex',23,TIMESTAMP '1970-01-01 00:00:01','par1'),
    ('id2','Stephen',33,TIMESTAMP '1970-01-01 00:00:02','par1'),
    ('id3','Julian',53,TIMESTAMP '1970-01-01 00:00:03','par2'),
    ('id4','Fabian',31,TIMESTAMP '1970-01-01 00:00:04','par2'),
    ('id5','Sophia',18,TIMESTAMP '1970-01-01 00:00:05','par3'),
    ('id6','Emma',20,TIMESTAMP '1970-01-01 00:00:06','par3'),
    ('id7','Bob',44,TIMESTAMP '1970-01-01 00:00:07','par4'),
    ('id8','Han',56,TIMESTAMP '1970-01-01 00:00:08','par4');

END;

  4. 将 Hudi SQL 脚本和 flink-sql-runner.jar 文件上传到 S3 位置。

  5. 在 FlinkDeployments YAML 文件中，将 hudi.enabled 设置为 true。

    spec:
  flinkConfiguration:
    hudi.enabled: "true"

  6. 创建一个 YAML 文件来运行配置。本示例文件名为 hudi-write.yaml。

    apiVersion: flink.apache.org/v1beta1
kind: FlinkDeployment
metadata:
  name: hudi-write-example
spec:
  flinkVersion: v1_18
  flinkConfiguration:
    taskmanager.numberOfTaskSlots: "2"
    hudi.enabled: "true"
  executionRoleArn: "<JobExecutionRole>"
  emrReleaseLabel: "emr-7.3.0-flink-latest"
  jobManager:
    highAvailabilityEnabled: false
    replicas: 1
    resource:
      memory: "2048m"
      cpu: 1
  taskManager:
    resource:
      memory: "2048m"
      cpu: 1
  job:
    jarURI: local:///opt/flink/usrlib/flink-sql-runner.jar
    args: ["/opt/flink/scripts/hudi-write.sql"]
    parallelism: 1
    upgradeMode: stateless
  podTemplate:
    spec:
      initContainers:
        - name: flink-sql-script-download
          args: 
            - s3
            - cp
            - s3://<s3_location>/hudi-write.sql
            - /flink-scripts
          image: amazon/aws-cli:latest
          imagePullPolicy: Always
          resources: {}
          terminationMessagePath: /dev/termination-log
          terminationMessagePolicy: File
          volumeMounts:
            - mountPath: /flink-scripts
              name: flink-scripts
        - name: flink-sql-runner-download
          args: 
            - s3
            - cp
            - s3://<s3_location>/flink-sql-runner.jar
            - /flink-artifacts
          image: amazon/aws-cli:latest
          imagePullPolicy: Always
          resources: {}
          terminationMessagePath: /dev/termination-log
          terminationMessagePolicy: File
          volumeMounts:
            - mountPath: /flink-artifacts
              name: flink-artifact
      containers:
        - name: flink-main-container
          volumeMounts:
            - mountPath: /opt/flink/scripts
              name: flink-scripts
            - mountPath: /opt/flink/usrlib
              name: flink-artifact
      volumes:
        - emptyDir: {}
          name: flink-scripts
        - emptyDir: {}
          name: flink-artifact

  7. 将 Flink Hudi 作业提交到 Flink Kubernetes Operator。

    kubectl apply -f hudi-write.yaml