使用tunnel同步PG数据到kafka
- 2019 年 10 月 26 日
- 笔记
tunnel同步PG数据到kafka
来自哈罗单车开源的组件。支持同步PG数据到kafka或者ES。
https://github.com/hellobike/tunnel
tunnel整体的部署比较简单的
需要事先部署好zk和kafka(我下面演示的是单节点的zk和kafka)
节点部署关系:
192.168.2.4 部署zk、kafka、pg10运行在1921端口
192.168.2.189 部署tunnel
确保已开启PG的逻辑复制
wal_level = 'logical'; max_replication_slots = 20
注意这个设置要重启PG进程的
然后,创建测试库表和同步用的账号
CREATE DATABASE test_database; c test_database create table test_1 (id int primary key , name char(40)); create table test_2 (id int primary key , name char(40)); CREATE ROLE test_rep LOGIN ENCRYPTED PASSWORD 'xxxx' REPLICATION; GRANT CONNECT ON DATABASE test_database to test_rep;
vim pg_hba.conf 增加2行配置:
host all test_rep 192.168.2.0/24 md5 host replication test_rep 192.168.2.0/24 md5
然后 reload 下PG
到192.168.2.189 机器上去 编译tunnel:
注意: tunnel的启动需要事先安装好oracle jdk 1.8
git clone https://github.com/hellobike/tunnel cd tunnel mvn clean package -Dmaven.test.skip=true cd target unzip AppTunnelService.zip cd AppTunnelService
vim conf/test.yml 内容如下:
tunnel_subscribe_config: pg_dump_path: '/usr/local/pgsql-10.10/bin/pg_dump' subscribes: - slotName: slot_for_test pgConnConf: host: 192.168.2.4 port: 1921 database: test_database user: test_rep password: xxxx rules: - {table: test_1, pks: ['id'], topic: test_1_logs} - {table: test_2, pks: ['id'], topic: test_2_logs} kafkaConf: addrs: - 192.168.2.4:9092 tunnel_zookeeper_address: 192.168.2.4:2181
前台启动:
java -server -classpath conf/*:lib/* com.hellobike.base.tunnel.TunnelLauncher -u false -c cfg.properties -p 7788 # 暴露prometheus metric在7788端口(配置监控不是这里的重点,也很简单,暂时先跳过)
然后,我们再在PG10上面的test_database的2张表随便造些数据,然后可以看到kafka里面已经有数据了(下图是通过kafkamanager和 kafka-eagle的结果)。


格式化下,数据就是这样的:
UPDATE的记录的样子:
{ "dataList": [{ "dataType": "integer", "name": "id", "value": "1111" }, { "dataType": "character", "name": "name", "value": "大狗蛋 " }], "eventType": "UPDATE", "lsn": 10503246616, "schema": "public", "table": "test_1" }
DELETE的记录的样子:
{ "dataList": [{ "dataType": "integer", "name": "id", "value": "3" }], "eventType": "DELETE", "lsn": 10503247064, "schema": "public", "table": "test_1" }