[ 
https://issues.apache.org/jira/browse/FLINK-35740?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Lee SeungMin updated FLINK-35740:
---------------------------------
    Description: 
PR: [Allow column as chunk key even if not in Primary Keys by SML0127 · Pull 
Request #3448 · apache/flink-cdc 
(github.com)|https://github.com/apache/flink-cdc/pull/3448]

Change the input chunk key to allow it even if it is not in the primary keys.

There are cases where the primary key is not a numeric type, such as varchar or 
varbinary. In this case, the `distributed factor` and `chunk range` may be 
calculated incorrectly, resulting in one chunk containing many records.

This may cause out of memory in the task manager.

Actually, in our company, there was a mysql tables that PK is a varbinary 
column and the above situation occurred.

  was:
Change the input chunk key to allow it even if it is not in the primary keys. 

There are cases where the primary key is not a numeric type, such as varchar or 
varbinary. In this case, the `distributed factor` and `chunk range` may be 
calculated incorrectly, resulting in one chunk containing many records.

This may cause out of memory in the task manager. 

Actually, in our company, there was a mysql tables that PK is a varbinary 
column and the above situation occurred.


> [mysql] Allow column as chunk key even if not in Primary Keys
> -------------------------------------------------------------
>
>                 Key: FLINK-35740
>                 URL: https://issues.apache.org/jira/browse/FLINK-35740
>             Project: Flink
>          Issue Type: Improvement
>          Components: Flink CDC
>            Reporter: Lee SeungMin
>            Priority: Major
>
> PR: [Allow column as chunk key even if not in Primary Keys by SML0127 · Pull 
> Request #3448 · apache/flink-cdc 
> (github.com)|https://github.com/apache/flink-cdc/pull/3448]
> Change the input chunk key to allow it even if it is not in the primary keys.
> There are cases where the primary key is not a numeric type, such as varchar 
> or varbinary. In this case, the `distributed factor` and `chunk range` may be 
> calculated incorrectly, resulting in one chunk containing many records.
> This may cause out of memory in the task manager.
> Actually, in our company, there was a mysql tables that PK is a varbinary 
> column and the above situation occurred.



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

Reply via email to