FlinkSql-cdc 实时ETL kafka 数据

以你之姓@ 2022-09-01 13:58 450阅读 0赞

文章目录

  • 资源参考
  • 1, 配置 debezium-mysql kafka 连接器
    • a, mysql开启binlog, 创建mysql 表和cdc用户
    • b, 使用 ksql 创建kafka连接器：debezium
    • c, flink sql 读写数据( mysql— > hbase)
  • 2, 配置 debezium-oracle kafka 连接器
    • a, oracle启用archive-log，并配置用户权限
    • b, 使用 ksql 创建kafka连接器：debezium
    • c, flink sql 读写数据 (oracle — > hbase)
  • 3, 使用 web 提交flink sql 任务
  • 4, flink 读取 confluent avro topic数据

资源参考

  • 参考云邪： https://github.com/ververica/flink-cdc-connectors
  • flink 社区：https://www.jianshu.com/u/504bc225cb7a，
    Bilibili视频（flink社区）：https://www.bilibili.com/video/BV1zt4y1D7kt/
  • flink join语法： https://ci.apache.org/projects/flink/flink-docs-release-1.13/zh/docs/dev/table/sql/queries/joins/index.html

1, 配置 debezium-mysql kafka 连接器

a, mysql开启binlog, 创建mysql 表和cdc用户

配置mysql: https://debezium.io/documentation/reference/1.3/connectors/mysql.html#setup-the-mysql-server

  1. mysql> show master status;
  2. +------------------+----------+--------------+------------------+
  3. | File             | Position | Binlog_Do_DB | Binlog_Ignore_DB |
  4. +------------------+----------+--------------+------------------+
  5. | mysql-bin.000001 | 10290780 |              |                  |
  6. +------------------+----------+--------------+------------------+
  7. 1 row in set (0.00 sec)
  9. mysql> use test2;
  10. mysql> show create table employee;
  11. +----------+--------------------------------------------------------------------                                                                                                                              
  12. | Table    | Create Table                                                                                                                                                                                                                                                                                                                 
  13. +----------+--------------------------------------------------------------------                                                               +
  14. | employee | CREATE TABLE `employee` (
  15.   `id` int(11) DEFAULT NULL,
  16.   `dept_id` int(11) DEFAULT NULL,
  17.   `name` varchar(20) COLLATE utf8_unicode_ci DEFAULT NULL
  18. ) ENGINE=InnoDB DEFAULT CHARSET=utf8 COLLATE=utf8_unicode_ci |
  19. +----------+--------------------------------------------------------------------
  20. 1 row in set (0.00 sec)

b, 使用 ksql 创建kafka连接器：debezium

  1. ksql> CREATE SOURCE CONNECTOR `mysql-dbz-src2` WITH(  
  2.      "connector.class"='io.debezium.connector.mysql.MySqlConnector',     
  3.      "tasks.max"='1',
  4.      "database.hostname"='127.0.0.1',
  5.      "database.port"='3306',
  6.      "database.server.name"='dbz2',
  8.      "database.user"='cdc',
  9.      "database.password"='cdc',
  11.      "database.whitelist"='test2',
  12.      "database.history.kafka.bootstrap.servers"='localhost:9092',
  13.      "database.history.kafka.topic"='schema-changes2.inventory2'
  14. );
  16. ksql> show connectors;
  17.  Connector Name  | Type   | Class                                         | Status
  18. --------------------------------------------------------------------------------------------------------
  19.  mysql-dbz-src2  | SOURCE | io.debezium.connector.mysql.MySqlConnector    | RUNNING (1/1 tasks RUNNING)
  20. --------------------------------------------------------------------------------------------------------
  22. ksql> show topics;
  23.  Kafka Topic                 | Partitions | Partition Replicas
  24. ---------------------------------------------------------------
  25.  dbz2                        | 1          | 1
  26.  dbz2.test2.dept             | 1          | 1
  27.  dbz2.test2.employee         | 1          | 1
  28.  dbz2.test2.enriched_orders  | 1          | 1
  29.  dbz2.test2.orders           | 1          | 1
  30.  dbz2.test2.products         | 1          | 1
  31.  dbz2.test2.salary           | 1          | 1
  32.  dbz2.test2.shipments        | 1          | 1
  33.  default_ksql_processing_log | 1          | 1
  34. ---------------------------------------------------------------

通过confluent platform查看kafka数据：
在这里插入图片描述
查看topic schema中：表字段和数据类型
在这里插入图片描述

c, flink sql 读写数据( mysql— > hbase)

  • dedebezium 连接器参数：https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/table/connectors/formats/debezium.html
  • hbase 连接器参数：https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/table/connectors/hbase.html

mysql join查询结果：（flink sql 执行类型sql来合并数据并写入hbase)
在这里插入图片描述

  1. [root@localhost bin]$ ./sql-client.sh "embedded"
  2.   |  ____| (_)     | |     / ____|/ __ \| |       / ____| (_)          | |
  3.    | |__  | |_ _ __ | | __ | (___ | |  | | |      | |    | |_  ___ _ __ | |_
  4.    |  __| | | | '_ \| |/ / \___ \| | | | | | | | | |/ _ \ '_ \| __|
  5.    | |    | | | | | |   <   ____) | |__| | |____  | |____| | |  __/ | | | |_
  6.    |_|    |_|_|_| |_|_|\_\ |_____/ \___\_\______|  \_____|_|_|\___|_| |_|\__|
  8.         Welcome! Enter 'HELP;' to list all available commands. 'QUIT;' to exit.
  11. Flink SQL>  // dbz2.test2.employee: (字段类型和 registry.url中保持一致，才能解析数据）
  12.         CREATE TABLE topic_test2_employee (
  13.           id int,
  14.           dept_id int,
  15.           name string
  16.         ) WITH (
  17.          'connector' = 'kafka',
  18.          'topic' = 'dbz2.test2.employee',
  19.          'properties.bootstrap.servers' = 'localhost:9092',
  20.          'scan.startup.mode' = 'earliest-offset',
  21.          'format' = 'debezium-avro-confluent',
  22.          'debezium-avro-confluent.schema-registry.url' = 'http://localhost:9081'
  23.         );
  25.         CREATE TABLE topic_test2_dept (
  26.               id int,
  27.               name string
  28.             ) WITH (
  29.              'connector' = 'kafka',
  30.              'topic' = 'dbz2.test2.dept',
  31.              'properties.bootstrap.servers' = 'localhost:9092',
  32.              'scan.startup.mode' = 'earliest-offset',
  33.              'format' = 'debezium-avro-confluent',
  34.              'debezium-avro-confluent.schema-registry.url' = 'http://localhost:9081'
  35.             );
  37.         CREATE TABLE topic_test2_salary (
  38.           id int,
  39.           emp_id int,
  40.           money double
  41.         ) WITH (
  42.          'connector' = 'kafka',
  43.          'topic' = 'dbz2.test2.salary',
  44.          'properties.bootstrap.servers' = 'localhost:9092',
  45.          'scan.startup.mode' = 'earliest-offset',
  46.          'format' = 'debezium-avro-confluent',
  47.          'debezium-avro-confluent.schema-registry.url' = 'http://localhost:9081'
  48.         );
  50.         CREATE TABLE hbase_www (
  51.          emp_id string,
  52.          f ROW<emp_dept_id string, name string , deptname string, salary string>,
  53.          PRIMARY KEY (emp_id) NOT ENFORCED
  54.         ) WITH (
  55.          'connector' = 'hbase-1.4',
  56.          'table-name' = 'www',
  57.          'zookeeper.quorum' = '192.168.56.117:2181'
  58.         );
  60. #关联查询
  61. insert into hbase_www
  62. select  cast(emp.id as string) emp_id, 
  63.     ROW(cast( emp.dept_id as string) , emp.name , dept.name , cast( salary.money as string)    )
  64. from topic_test2_employee  emp
  65. inner join topic_test2_dept dept on dept.id=emp.dept_id
  66. inner join topic_test2_salary salary on salary.emp_id= emp.id
  68. [INFO] Submitting SQL update statement to the cluster...
  69. [INFO] Table update statement has been successfully submitted to the cluster:
  70. Job ID: 137c51550417ab7003c00c95d74f842b

登录flink web ui: 查看运行的任务
在这里插入图片描述查看hbase 数据是否实时变化（mysql insert/update/delete数据后，是否同步变化）

  1. hbase(main):016:0> scan 'xx_t1'
  2. ROW                                      COLUMN+CELL
  3.  1                                       column=f:dept_id, timestamp=1627454335726, value=1
  4.  1                                       column=f:name, timestamp=1627454335726, value=Li si
  5.  2                                       column=f:dept_id, timestamp=1627454335726, value=2
  6.  2                                       column=f:name, timestamp=1627454335726, value=Test 1
  7. 2 row(s) in 0.0120 seconds

2, 配置 debezium-oracle kafka 连接器

https://debezium.io/documentation/reference/1.6/connectors/oracle.html#oracle-deploying-a-connector
oracle join 查询的数据：
在这里插入图片描述

a, oracle启用archive-log，并配置用户权限

  1. SQL> shutdown immediate    
  2.     startup mount
  3.     alter database archivelog;
  4.     alter database open;
  5.     -- Should now "Database log mode: Archive Mode"
  6.     -- archive log list 
  7.     ALTER TABLE user1.table1 ADD SUPPLEMENTAL LOG DATA (ALL) COLUMNS;
  8.     ALTER DATABASE ADD SUPPLEMENTAL LOG DATA;
  10. SQL> CREATE TABLESPACE logminer_tbs DATAFILE '/home/oracle/logminer_tbs.dbf' SIZE 25M REUSE AUTOEXTEND ON MAXSIZE UNLIMITED;
  11.     CREATE USER dbz IDENTIFIED BY dbz DEFAULT TABLESPACE logminer_tbs   QUOTA UNLIMITED ON logminer_tbs ;
  13.     GRANT CREATE SESSION TO dbz ;
  14.     GRANT SELECT ON V_$DATABASE to dbz ;
  15.     GRANT FLASHBACK ANY TABLE TO dbz ;
  16.     GRANT SELECT ANY TABLE TO dbz ;
  17.     GRANT SELECT_CATALOG_ROLE TO dbz ;
  18.     GRANT EXECUTE_CATALOG_ROLE TO dbz ;
  19.     GRANT SELECT ANY TRANSACTION TO dbz ;    
  20.     --GRANT LOGMINING TO dbz ; 11g err: 替换为以下两行
  21.     GRANT EXECUTE ON SYS.DBMS_LOGMNR TO dbz;
  22.     GRANT SELECT ON V_$LOGMNR_CONTENTS TO dbz;
  23.     GRANT CREATE TABLE TO dbz ;
  24.     GRANT LOCK ANY TABLE TO dbz ;
  25.     GRANT ALTER ANY TABLE TO dbz ;
  26.     GRANT CREATE SEQUENCE TO dbz ;
  28.     GRANT EXECUTE ON DBMS_LOGMNR TO dbz ;
  29.     GRANT EXECUTE ON DBMS_LOGMNR_D TO dbz ;    
  30.     GRANT SELECT ON V_$LOG TO dbz ;
  31.     GRANT SELECT ON V_$LOG_HISTORY TO dbz ;
  32.     GRANT SELECT ON V_$LOGMNR_LOGS TO dbz ;    
  33.     GRANT SELECT ON V_$LOGMNR_CONTENTS TO dbz ;
  34.     GRANT SELECT ON V_$LOGMNR_PARAMETERS TO dbz ;
  35.     GRANT SELECT ON V_$LOGFILE TO dbz ;
  36.     GRANT SELECT ON V_$ARCHIVED_LOG TO dbz ;
  37.     GRANT SELECT ON V_$ARCHIVE_DEST_STATUS TO dbz ;

b, 使用 ksql 创建kafka连接器：debezium

  • The Debezium Oracle connector by default ingests changes using native Oracle LogMiner.

    ksql> CREATE SOURCE CONNECTOR oracle_cdc4 WITH( 

    1.    "connector.class"='io.debezium.connector.oracle.OracleConnector',     
    2.    "tasks.max" ='1',
    3.     "database.server.name" ='helowin_cdc4',
    4.     "database.hostname" ='localhost',
    5.     "database.port" ='15211',
    6.     "database.user" ='dbz',
    7.     "database.password" ='dbz',
    8.     "schema.include.list"='SCOTT',
    9.     "database.dbname" ='helowin',
    10.     "decimal.handling.mode"='string',
    11.     "database.history.kafka.bootstrap.servers" ='localhost:9092',
    12.     "database.history.kafka.topic"='schema-changes.oracle_cdc4',
    13.     "database.history.skip.unparseable.ddl"=true

    );

    ksql> show connectors;

    Connector Name | Type | Class | Status

    oracle_cdc4 | SOURCE | io.debezium.connector.oracle.OracleConnector | RUNNING (1/1 tasks RUNNING)

    ksql> show topics;

    Kafka Topic | Partitions | Partition Replicas

    helowin_cdc4 | 1 | 1
    helowin_cdc4.SCOTT.DEPT | 1 | 1
    helowin_cdc4.SCOTT.EMP | 1 | 1
    helowin_cdc4.SCOTT.SALGRADE | 1 | 1
    helowin_cdc4.SCOTT.TEST | 1 | 1

    ksql> set ‘auto.offset.reset’=’earliest’;
    Successfully changed local property ‘auto.offset.reset’ from ‘earliest’ to ‘earliest’.
    ksql> print helowin_cdc4.SCOTT.TEST;
    Key format: SESSION(AVRO) or HOPPING(AVRO) or TUMBLING(AVRO) or AVRO or SESSION(PROTOBUF) or HOPPING(PROTOBUF) or TUMBLING(PROTOBUF) or PROTOBUF or SESSION(JSON) or HOPPING(JSON) or TUMBLING(JSON) or JSON or SESSION(JSON_SR) or HOPPING(JSON_SR) or TUMBLING(JSON_SR) or JSON_SR or SESSION(KAFKA_INT) or HOPPING(KAFKA_INT) or TUMBLING(KAFKA_INT) or KAFKA_INT or SESSION(KAFKA_BIGINT) or HOPPING(KAFKA_BIGINT) or TUMBLING(KAFKA_BIGINT) or KAFKA_BIGINT or SESSION(KAFKA_DOUBLE) or HOPPING(KAFKA_DOUBLE) or TUMBLING(KAFKA_DOUBLE) or KAFKA_DOUBLE or SESSION(KAFKA_STRING) or HOPPING(KAFKA_STRING) or TUMBLING(KAFKA_STRING) or KAFKA_STRING
    Value format: AVRO
    rowtime: 7/29/21 11:33:09 AM CST, key: , value: { “before”: null, “after”: { “ID”: “2”, “NAME”: “b”, “MARK”: “b”}, “source”: { “version”: “1.6.1.Final”, “connector”: “oracle”, “name”: “helowin_cdc4”, “ts_ms”: 1627529588336, “snapshot”: “last”, “db”: “HELOWIN”, “sequence”: null, “schema”: “SCOTT”, “table”: “TEST”, “txId”: null, “scn”: “1103657”, “commit_scn”: null, “lcr_position”: null}, “op”: “r”, “ts_ms”: 1627529588336, “transaction”: null}
    rowtime: 7/29/21 12:11:18 PM CST, key: , value: { “before”: null, “after”: { “ID”: “3”, “NAME”: “c”, “MARK”: “c”}, “source”: { “version”: “1.6.1.Final”, “connector”: “oracle”, “name”: “helowin_cdc4”, “ts_ms”: 1627511996000, “snapshot”: “false”, “db”: “HELOWIN”, “sequence”: null, “schema”: “SCOTT”, “table”: “TEST”, “txId”: “02001f0051030000”, “scn”: “1145746”, “commit_scn”: “1145747”, “lcr_position”: null}, “op”: “c”, “ts_ms”: 1627531878417, “transaction”: null}
    rowtime: 7/29/21 12:12:07 PM CST, key: , value: { “before”: { “ID”: “3”, “NAME”: “c”, “MARK”: “c”}, “after”: { “ID”: “3”, “NAME”: “c2”, “MARK”: “c2”}, “source”: { “version”: “1.6.1.Final”, “connector”: “oracle”, “name”: “helowin_cdc4”, “ts_ms”: 1627512679000, “snapshot”: “false”, “db”: “HELOWIN”, “sequence”: null, “schema”: “SCOTT”, “table”: “TEST”, “txId”: “0200080068030000”, “scn”: “1145746”, “commit_scn”: “1174275”, “lcr_position”: null}, “op”: “u”, “ts_ms”: 1627531927799, “transaction”: null}

c, flink sql 读写数据 (oracle – > hbase)

  1. Flink SQL> CREATE TABLE emp (
  2.               EMPNO int, ENAME string, JOB string, MGR int, HIREDATE BIGINT , SAL string , COMM string, DEPTNO int
  3.             ) WITH (
  4.               'connector' = 'kafka',
  5.               'properties.bootstrap.servers' = 'localhost:9092',
  6.               'topic' = 'helowin_cdc4.SCOTT.EMP',
  7.               'format' = 'debezium-avro-confluent',
  8.               'debezium-avro-confluent.schema-registry.url' = 'http://localhost:9081',
  9.               'scan.startup.mode' = 'earliest-offset'
  10.             );
  12.             CREATE TABLE dept (
  13.               DEPTNO int, DNAME string, LOC string
  14.             ) WITH (
  15.               'connector' = 'kafka',
  16.               'properties.bootstrap.servers' = 'localhost:9092',
  17.               'topic' = 'helowin_cdc4.SCOTT.DEPT',
  18.               'format' = 'debezium-avro-confluent',
  19.               'debezium-avro-confluent.schema-registry.url' = 'http://localhost:9081',
  20.               'scan.startup.mode' = 'earliest-offset'
  21.             );
  23.             CREATE TABLE hbase_www2 (
  24.              EMPNO   string ,
  25.              f ROW<ENAME string, JOB string , MGR string, 
  26.                  HIREDATE string, SAL string, COMM string, 
  27.                  DEPTNO string, DNAME string,  LOC string>,
  28.              PRIMARY KEY (EMPNO ) NOT ENFORCED
  29.             ) WITH (
  30.              'connector' = 'hbase-1.4',
  31.              'table-name' = 'www',
  32.              'zookeeper.quorum' = 'localhost:2181'
  33.             );
  35. //合并多个表的flink 视图
  36. create view dept_emp_v as
  37. select cast(EMPNO as string) EMPNO, ENAME , JOB, cast(MGR as string) MGR, cast(HIREDATE as string) HIREDATE, SAL, COMM, cast(emp.DEPTNO as string) DEPTNO, dept.DNAME ,dept.LOC 
  38. from  emp
  39. INNER JOIN dept 
  40. ON emp.DEPTNO =dept.DEPTNO 
  42. //从视图中检索数据并插入hbase
  43. insert into hbase_www2
  44. select EMPNO , ROW(  ENAME , JOB, MGR , HIREDATE, SAL, COMM, DEPTNO , DNAME ,LOC   )
  45. from  dept_emp_v
  47. [INFO] Submitting SQL update statement to the cluster...
  48. [INFO] Table update statement has been successfully submitted to the cluster:
  49. Job ID: e73a8f4c8860a3bc83879e1a50543350

3, 使用 web 提交flink sql 任务

https://gitee.com/zhuhuipei/flink-streaming-platform-web/blob/master/docs/deploy.md
登录web ui: http://$\{ip或者hostname\}:9084/ , admin 123456 (编辑sql任务）

  • 注意：创建任务时，三方jar地址 那一栏最好留空，把相关jar包手动下载到 flink_home/lib, 排除http://ccblog.cn/jars/flink-streaming-udf.jar 这个jar包避免和flink内置的函数有冲突

  • platform 工作原理：本质是调用了flink客户端，提交一个jar包任务
    flink run -d -p 2
    -c com.flink.streaming.core.JobApplication
    flink-streaming-platform-web/lib/flink-streaming-core_flink_1.12.0-1.2.0.RELEASE.jar
    -sql flink-streaming-platform-web/sql/job_sql_8.sql

    常用jar包

    https://ci.apache.org/projects/flink/flink-docs-release-1.12/dev/table/connectors/formats/avro-confluent.html
    https://repo1.maven.org/maven2/com/alibaba/ververica/flink-sql-connector-mysql-cdc/1.1.0/flink-sql-connector-mysql-cdc-1.1.0.jar
    https://repo.maven.apache.org/maven2/org/apache/flink/flink-sql-connector-hbase-1.4_2.11/1.12.3/flink-sql-connector-hbase-1.4_2.11-1.12.3.jar
    http://ccblog.cn/jars/flink-connector-jdbc_2.11-1.12.0.jar
    http://ccblog.cn/jars/flink-sql-connector-kafka_2.11-1.12.0.jar

在这里插入图片描述
在这里插入图片描述

在这里插入图片描述

4, flink 读取 confluent avro topic数据

在这里插入图片描述

  1. ksql> CREATE STREAM t1 (id int ,  name VARCHAR) WITH ( KAFKA_TOPIC='t1', PARTITIONS='2',VALUE_FORMAT='avro');
  2. ksql> insert into t1(id,name) values (1,'a');
  3. ksql> select * from t1 emit changes;
  4. +------------------+------------------+------------------+------------------+
  5. |ROWTIME           |ROWKEY            |ID                |NAME              |
  6. +------------------+------------------+------------------+------------------+
  7. |1627631954797     |null              |1                 |a                 |
  10. Flink SQL> CREATE TABLE t1 (
  11.               ID int,
  12.               NAME string
  13.             ) WITH (
  14.               'connector' = 'kafka',
  15.               'properties.bootstrap.servers' = 'localhost:9092',
  16.               'topic' = 't1',
  17.               'format' = 'avro-confluent',
  18.               'avro-confluent.schema-registry.url' = 'http://localhost:9081',
  19.               'scan.startup.mode' = 'earliest-offset'
  20.             );

发表评论

表情：
评论列表 （有 0 条评论，450人围观）

还没有评论，来说两句吧...

相关阅读

    相关 ETL数据清洗

    ![45197bba580040bfb64f3e7562cc9ea4.png][] 大多[数据仓库][Link 1]的数据架构可以概括为： 数据源-->ODS(操作型数据存

    小鱼儿小鱼儿/ 0 赞/ 139 阅读

    相关 Flink：实时ETL案例

    抽取（extract）、转换（transform）、加载（load） 需求 将国家编号变成地区编号（在跨国业务常用） 形式 数据源 \{"dt":"2020-

    男娘i男娘i/ 0 赞/ 108 阅读