作者 | 番兄
如何借助 Apache 将 MySQL 数据高效同步至 S3file？本文详述的步骤已全部通过测试验证，适用于构建基于对象存储的数据中台场景，具备部署灵活、扩展性强等优势，对有 MySQL 到 S3 数据集成需求的用户具有较高的参考价值，点赞、收藏学习吧！
第一步：创建Hive表
CREATE EXTERNAL TABLE ods_ekp.`ods_sys_notify_todo_bak` ( `fd_id` STRING, `fd_app_name` STRING, `fd_model_name` STRING, `fd_model_id` STRING, `fd_key` STRING, `fd_parameter1` STRING, `fd_parameter2` STRING, `fd_create_time` TIMESTAMP, `fd_subject` STRING, `fd_type` INT, `fd_link` STRING, `fd_mobile_link` STRING, `fd_pad_link` STRING, `fd_bundle` STRING, `fd_replace_text` STRING, `fd_md5` STRING, `fd_del_flag` STRING, `fd_level` INT, `doc_creator_id` STRING, `fd_extend_content` STRING, `fd_lang` STRING, `fd_cate_name` STRING, `fd_cate_id` STRING, `fd_template_name` STRING, `fd_template_id` STRING, `fd_hierarchy_id` STRING)COMMENT 'sys_notify_todo_bak data'PARTITIONED BY ( `dctime` STRING COMMENT '分区年月日')ROW FORMAT DELIMITED FIELDS TERMINATED BY '\001'STORED AS PARQUETLOCATION 's3a://seatunnel/doris/warehouse/ods_ekp/ods_sys_notify_todo_bak'TBLPROPERTIES ( 'parquet.compression'='ZSTD');
注意：
ROW FORMAT DELIMITED FIELDS TERMINATED BY '\001' 这个分隔符设置需要在后面SeaTunnel里面配置一样的，不然格式错误;
'parquet.compression'='ZSTD' 压缩算法也是需要在后面SeaTunnel里面配置一样的;
STORED AS PARQUET parquet文件格式，也是需要在后面SeaTunnel里面配置一样的.
是用之前把注释删除
env { job.mode = "BATCH" parallelism = 2}source { Jdbc { url = "jdbc:mysql://[服务器ip]:3306/[数据库]?serverTimezone=GMT%2b8&useUnicode=true&characterEncoding=UTF-8&rewriteBatchedStatements=true" driver = "com.mysql.cj.jdbc.Driver" user = "[账户]" password = "[密码]", # dctime要转换成字符串，因为hive建表的时候，这个字段是字符串；分区字段也要加到查询里面，后面SeaTunnel sink的时候会自动处理 query = "select fd_id, fd_app_name, fd_model_name, fd_model_id, fd_key, fd_parameter1, fd_parameter2, fd_create_time, fd_subject, fd_type, fd_link, fd_mobile_link, fd_pad_link, fd_bundle, fd_replace_text, fd_md5, fd_del_flag, fd_level, doc_creator_id, fd_extend_content, fd_lang, fd_cate_name, fd_cate_id, fd_template_name, fd_template_id, fd_hierarchy_id, cast(date_format(fd_create_time, '%Y-%m-%d') as char) as dctime from sys_notify_todo_bak }}transform {}sink { S3File { bucket = "s3a://seatunnel" fs.s3a.endpoint = "[minio服务器域名/ip]:9000" access_key = "[账户]" secret_key = "[密码]" fs.s3a.aws.credentials.provider="org.apache.hadoop.fs.s3a.SimpleAWSCredentialsProvider" # 目录地址 path = "/doris/warehouse/ods_ekp/ods_sys_notify_todo_bak" tmp_path = "/data/tmp/seatunnel" # 必须填写的不然会出现问题，因为本人的minio没有做ssl处理，所以要这么设置 hadoop_s3_properties { "fs.s3a.connection.ssl.enabled" = "false" "fs.s3a.path.style.access" = "true" } # parquet文件格式 file_format_type = "parquet" # 必须用\\代表\ field_delimiter = "\\001" # parquet格式必须要加，否则会出问题 parquet_avro_write_timestamp_as_int96 = true # 压缩算法 compress_codec = "zstd" have_partition = true partition_by = ["dctime"] partition_dir_expression = "${k0}=${v0}" is_partition_field_write_in_file = false schema_save_mode = "CREATE_SCHEMA_WHEN_NOT_EXIST" data_save_mode="APPEND_DATA" custom_filename = true file_name_expression = "${transactionId}_${now}" filename_time_format = "yyyy.MM.dd" }}第二步：执行SeaTunnel同步，并在Hive查询里执行下列操作
-- 刷新物理目录分析MSCK REPAIR TABLE ods_ekp.ods_sys_notify_todo_bak;-- 查询hive表确认是否有数据SELECT * from ods_ekp.ods_sys_notify_todo_bak limit 100;
第三步：创建Doris Hive catalog外部库
CREATE CATALOG hive PROPERTIES ( 'type'='hms', 'hive.metastore.uris' = 'thrift://[hive metastore server的ip]:9083', "s3.endpoint" = "http://[minio服务器域名/ip]:9000", "s3.region" = "us-east-1", "s3.access_key" = "[账户]", "s3.secret_key" = "[密码]", "s3.connection.ssl.enabled" = "false", "use_path_style" = "true", "hive.version" = '2.1.1');REFRESH CATALOG hive;show databases from hive;SELECT * from hive.ods_ekp.ods_sys_notify_todo_bak limit 100
说明：
因为本人用的CDH6.3.2版本，Hive是2.1.1版本，所以建立catalog的时候，需要指定"hive.version" = '2.1.1'。
因为本人设置的minio没有ssl，所以配置的时候需要加上"s3.connection.ssl.enabled" = "false"。
Minio用的是path风格，所以需要配置"use_path_style" = "true"。
SeaTunnel版本： 2.3.11
Doris版本：2.0.15
