[Bug] [dinky-metadata-base] MySQL wait_timeout After closing the idle connection, the metadata data source connection permanently fails and needs to be restarted to recover. #4570

Search before asking

• I had searched in the issues and found no similar issues.

What happened

在 Dinky 中注册的 MySQL 数据源,如果空闲时间超过 MySQL 服务端的 wait_timeout(默认 8 小时),
下次通过该数据源执行操作(如提交 CDCSOURCE 任务读取表元数据)时会抛出 CommunicationsException,
且该数据源之后一直不可用,只能重启 Dinky 才能恢复。

What you expected to happen

Driver 应能检测到失效连接,并自动重建连接,业务无感知。

How to reproduce

1. 在 Dinky(1.2.5)中注册一个 MySQL 数据源
2. 让该数据源空闲超过 MySQL 的 wait_timeout (快速复现可在 MySQL 服务端设置 SET GLOBAL wait_timeout=60)
3. 提交一个通过该数据源读取元数据的任务(如 CDCSOURCE 作业)
   或者提交了一个CDCSOURCE 作业;第二天重启报错复现

Anything else

ERROR com.alibaba.druid.pool.DruidPooledStatement: CommunicationsException, druid version 1.2.23, jdbcUrl : jdbc:mysql://...:3306/test?..., testWhileIdle true, idle millis 58786359, minIdle 5, poolingCount 0 ... ERROR com.alibaba.druid.pool.DruidDataSource: {conn-310002} discard com.mysql.cj.jdbc.exceptions.CommunicationsException: The last packet successfully received from the server was 58,786,341 milliseconds ago ... is longer than the server configured value of 'wait_timeout' ...

Version

dev

Are you willing to submit PR?

• Yes I am willing to submit a PR!

Code of Conduct

• I agree to follow this project's Code of Conduct