Flink Postgres CDC

大数据技术AI

Flink/Spark/Hadoop/数仓，数据分析、面试，源码解读等干货学习资料

128篇原创内容

公众号

1、Flink JDBC 写 Postgres

添加依赖

 
     org.apache.flink
     flink-connector-jdbc_2.11
     ${flink.version}
 

代码

import org.apache.flink.configuration.Configuration;
 import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
 import org.apache.flink.table.api.EnvironmentSettings;
 import org.apache.flink.table.api.bridge.java.StreamTableEnvironment;

 public class Flink_JDBC_PG {
     public static void main(String[] args) {

         Configuration conf = new Configuration();
         StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment(conf);
         EnvironmentSettings settings = EnvironmentSettings
                 .newInstance()
                 .inStreamingMode()
                 .build();

         StreamTableEnvironment tableEnvironment = StreamTableEnvironment.create(env, settings);

         String source = "CREATE TABLE student (\n" +
                 "    id INT,\n" +
                 "    age  INT,\n" +
                 "    name   STRING,\n" +
                 "    create_time   TIMESTAMP(3),\n" +
                 "    WATERMARK FOR create_time AS create_time\n" +
                 ") WITH (\n" +
                 "  'connector' = 'datagen',\n" +
                 "  'rows-per-second'='20',\n" +
                 "  'fields.id.kind'='random' ,\n" +
                 "  'fields.id.min'='1',\n" +
                 "  'fields.id.max'='100',\n" +
                 "  'fields.age.kind'='random',\n" +
                 "  'fields.age.min'='1',\n" +
                 "  'fields.age.max'='100',\n" +
                 "  'fields.name.kind'='random',\n" +
                 "  'fields.name.length'='3'\n" +
                 ")";

         tableEnvironment.executeSql(source);

         String sink_sql = "CREATE TABLE sink_student (\n" +
                 " id INT,\n" +
                 " age  INT,\n" +
                 " name   STRING,\n" +
                 " PRIMARY KEY (id) NOT ENFORCED\n" +
                 ") WITH (\n" +
                 "   'connector' = 'jdbc',\n" +
                 "   'url' = 'jdbc:postgresql://bigdata:5432/postgres',\n" +
                 "   'table-name' = 'student',\n" +
                 "   'username'='postgres',\n" +
                 "   'password'='123456'\n" +
                 ")";

         tableEnvironment.executeSql(sink_sql);
         String result = "insert into sink_student select id,age,name from student";
         tableEnvironment.executeSql(result);
         tableEnvironment.executeSql("select * from student").print();

     }
 }

pg表要提前创建好，要有主键

2、Postgres 配置binlog

2.1 更改配置文件postgresql.conf

# 更改wal日志方式为logical
 wal_level = logical # minimal, replica, or logical
 # 更改solts最大数量（默认值为10），flink-cdc默认一张表占用一个slots
 max_replication_slots = 20 # max number of replication slots
 # 更改wal发送最大进程数（默认值为10），这个值和上面的solts设置一样
 max_wal_senders = 20 # max number of walsender processes
 # 中断那些停止活动超过指定毫秒数的复制连接，可以适当设置大一点（默认60s）
 wal_sender_timeout = 180s # in milliseconds; 0 disable　

重新加载配置或重启postgres

 [root@bigdata ~]# docker exec -it 7b2d8a96ef4c /bin/bash
 root@7b2d8a96ef4c:/# su postgres
 postgres@7b2d8a96ef4c:/$ pg_ctl reload
 server signaled

查看参数

 select *
 from pg_settings
 where name in ('wal_level',
                'max_replication_slots',
                'max_wal_senders',
                'wal_sender_timeout');

2.2 给用户复制流权限

 -- 给用户复制流权限
 ALTER ROLE postgres REPLICATION;
 -- 给用户登录数据库权限
 GRANT CONNECT ON DATABASE postgres TO postgres;
 -- 把当前库public下所有表查询权限赋给用户
 GRANT SELECT ON ALL TABLES IN SCHEMA public TO postgres;

2.3 发布表

 --  更新发布所有表标识（可选）
 UPDATE pg_publication SET puballtables= false WHERE pubname IS NOT NULL;
 -- 查看发布设置
 SELECT * FROM pg_publication;
 -- 查看那些表发布
 SELECT * FROM pg_publication_tables;
 -- 创建发布所有表
 CREATE PUBLICATION dbz_publication FOR ALL TABLES;
 -- 为发布添加一个表
 ALTER PUBLICATION dbz_publication ADD TABLE student;

2.4 更改表的复制标识包含更新和删除的值

 -- 更改复制标识包含更新和删除之前值
 ALTER TABLE student REPLICA IDENTITY FULL;
 -- 查看复制标识（为f标识说明设置成功）
 SELECT relreplident FROM pg_class WHERE relname = 'student';

2.5 常用pg命令

-- 可以不用重启postgres cluster 就可以生效配置
pg_ctl reload


-- 创建用户
CREATE USER root WITH PASSWORD '123456';
-- 给用户增删改查权限
GRANT INSERT, UPDATE, SELECT, DELETE ON ALL TABLES IN SCHEMA public TO postgres;
-- 给用户复制流权限
ALTER ROLE postgres REPLICATION;
-- 给用户登录数据库权限
GRANT CONNECT ON DATABASE postgres TO postgres;
-- 把当前库public下所有表查询权限赋给用户
GRANT SELECT ON ALL TABLES IN SCHEMA public TO postgres;


UPDATE pg_publication SET puballtables= false WHERE pubname IS NOT NULL;
-- 查看发布设置
SELECT * FROM pg_publication;
-- 查看那些表发布
SELECT * FROM pg_publication_tables;
-- 创建发布所有表
CREATE PUBLICATION dbz_publication FOR ALL TABLES;
-- 为发布添加一个表
ALTER PUBLICATION dbz_publication ADD TABLE student;

-- 更改复制标识包含更新和删除之前值
ALTER TABLE student REPLICA IDENTITY FULL;
-- 查看复制标识（为f标识说明设置成功）
SELECT relreplident FROM pg_class WHERE relname = 'student';


-- 删除slot
SELECT PG_DROP_REPLICATION_SLOT('my_slot');
-- 查询用户连接数
SELECT usename, COUNT(*) FROM pg_stat_activity GROUP BY usename ORDER BY COUNT(*) DESC;
-- 设置用户最大连接数
ALTER ROLE postgres CONNECTION LIMIT 200;

3、Flink Streaming Postgres CDC

3.1 引入依赖

    com.ververica
    flink-sql-connector-postgres-cdc
    2.1.0

3.2 代码

import com.ververica.cdc.connectors.postgres.PostgreSQLSource;
 import com.ververica.cdc.debezium.JsonDebeziumDeserializationSchema;
 import org.apache.flink.streaming.api.datastream.DataStreamSource;
 import org.apache.flink.streaming.api.environment.StreamExecutionEnvironment;
 import org.apache.flink.streaming.api.functions.source.SourceFunction;

 public class FlinkStream_CDC_PG {
     public static void main(String[] args) throws Exception {

         StreamExecutionEnvironment env = StreamExecutionEnvironment.getExecutionEnvironment();

         SourceFunction sourceFunction = PostgreSQLSource.builder()
                 .hostname("bigdata")
                 .port(5432)
                 .database("postgres")
                 .schemaList("public")
                 .tableList("public.student")
                 .username("postgres")
                 .password("123456")
                 .slotName("sink_student_cdc1")
                 .decodingPluginName("pgoutput")
                 .deserializer(new JsonDebeziumDeserializationSchema())
                 .build();
         DataStreamSource streamSource = env.addSource(sourceFunction);
         streamSource.print();

         env.execute();

     }
 }

4、Flink SQL Postgres CDC

 import org.apache.flink.table.api.EnvironmentSettings;
 import org.apache.flink.table.api.TableEnvironment;

 public class FlinkSql_CDC_PG {
     public static void main(String[] args) {

         EnvironmentSettings settings = EnvironmentSettings
                 .newInstance()
                 .inStreamingMode()
                 .useBlinkPlanner()
                 .build();

         TableEnvironment tableEnvironment = TableEnvironment.create(settings);

         tableEnvironment.getConfig().getConfiguration().setString("execution.checkpointing.interval", "3s");

         String pg_sql = "CREATE TABLE sink_student_cdc1 (\n" +
                 " id INT,\n" +
                 " age  INT,\n" +
                 " name   STRING\n" +
                 ") WITH (\n" +
                 "  'connector' = 'postgres-cdc',\n" +
                 "  'hostname' = '106.52.242.238',\n" +
                 "  'port' = '5432',\n" +
                 "  'username' = 'postgres',\n" +
                 "  'password' = '123456',\n" +
                 "  'database-name' = 'postgres',\n" +
                 "  'schema-name' = 'public',\n" +
                 "  'table-name' = 'student',\n" +
                 "  'decoding.plugin.name' = 'pgoutput',\n" +
                 "  'slot.name'='sink_student_cdc1',\n" +
                 "  'debezium.snapshot.mode' = 'exported'\n" +
                 ")";
         tableEnvironment.executeSql(pg_sql);
         tableEnvironment.executeSql("select * from sink_student_cdc1").print();

     }
 }

5、问题汇总

Q1：PSQLException: ERROR: replication slot “flink” is active for PID 974 error

slot.name 要唯一，一张表一个

Q2：使用CDC 2.x版本，只能读取全量数据，无法读取增量（binlog） 数据

CDC 2.0 支持了无锁算法，支持并发读取，为了保证全量数据 + 增量数据的顺序性，依赖Flink 的 checkpoint机制，所以作业需要配置 checkpoint。SQL 作业中配置方式：

 Flink SQL> SET 'execution.checkpointing.interval' = '3s';

DataStream 作业配置方式：

 env.enableCheckpointing(3000);

PGCDC：更改表的复制标识包含更新和删除

Q3：作业报错 Replication slot “xxxx” is active

• 前往 Postgres 中手动执行以下命令

 select pg_drop_replication_slot('rep_slot');

• pg source with 参数中添加 'debezium.slot.drop.on.stop' = 'true'，在作业停止后自动清理 slot

Q4：什么时候使用 flink-sql-connector-xxx.jar，什么时候使用 flink-connector-xxx.jar，两者有啥区别?

Flink CDC 项目中各个connector的依赖管理和Flink 项目中 connector 保持一致。

flink-sql-connector-xx 是胖包，除了connector的代码外，还把 connector 依赖的所有三方包 shade 后打入，提供给 SQL 作业使用，用户只需要在 lib目录下添加该胖包即可。

flink-connector-xx 只有该 connector 的代码，不包含其所需的依赖，提供 datastream 作业使用，用户需要自己管理所需的三方包依赖，有冲突的依赖需要自己做 exclude, shade 处理。

Q5：decoding.plugin.name 无法访问文件 “decoderbufs”: 没有那个文件或目录

根据Postgres服务上安装的插件确定。支持的插件列表如下：

• decoderbufs（默认值）

• wal2json

• wal2json_rds

• wal2json_streaming

• wal2json_rds_streaming

• pgoutput

Q6：SlotName

Slot names must conform to PostgreSQL replication slot naming rules, which state: “Each replication slot has a name, which can contain lower-case letters, numbers, and the underscore character.”

Q7：snapshot.mode

initial- 只有当逻辑服务器名称没有记录偏移量时，连接器才会执行快照。

always- 每次启动连接器时，连接器都会执行快照。

never- 连接器从不执行快照。当连接器以这种方式配置时，其启动时的行为如下。如果卡夫卡偏移量主题中之前存储的LSN，连接器将继续从该位置流式传输更改。如果没有存储LSN，连接器将从服务器上创建PostgreSQL逻辑复制插槽的时间点开始流式传输更改。只有当您知道所有感兴趣的数据仍然反映在WAL中时，never快照模式才有用。

initial_only-连接器执行初始快照，然后停止，而不处理任何后续更改。

exported- 连接器根据创建复制插槽的时间点执行快照。这是以无锁方式执行快照的绝佳方式。