You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@flink.apache.org by "Arsene (Jira)" <ji...@apache.org> on 2025/05/23 00:24:00 UTC

[jira] [Created] (FLINK-37834) Flink JDBC support ClickHouse Dialect

Arsene created FLINK-37834:
------------------------------

             Summary: Flink JDBC support ClickHouse Dialect
                 Key: FLINK-37834
                 URL: https://issues.apache.org/jira/browse/FLINK-37834
             Project: Flink
          Issue Type: New Feature
          Components: Connectors / JDBC, Table SQL / API
    Affects Versions: jdbc-4.0.0, jdbc-3.2.0
            Reporter: Arsene


The latest version of the Flink JDBC connector does not support the ClickHouse dialect, so starting a streaming job to ClickHouse using the Flink SQL CLI is not currently supported.

 

There is an opened PR [https://github.com/apache/flink-connector-jdbc/pull/49] yet that version has some issues and does not support array/map types for ClickHouse.

 

There is also an open-source third-party ClickHouse connector available at [https://github.com/itinycheng/flink-connector-clickhouse]. However, it has a major issue: if the Flink job encounters an error after a successful start—such as an {{SQLException}} due to a failure to write to the database or if the target ClickHouse table is dropped after the job has started—the exception is only logged in the TaskManager. As a result, the job does not restart, no checkpoint is triggered, and this can lead to data loss.

 

I have a working ClickHouse dialect implementation for {{flink-jdbc-connector}} version 3.2.0, and I'm interested in contributing it to the Flink project. I plan to update the implementation to be compatible with the latest version so that ClickHouse support can be officially included in the JDBC connector distribution.



--
This message was sent by Atlassian Jira
(v8.20.10#820010)