[
https://issues.apache.org/jira/browse/FLINK-38235?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=18013565#comment-18013565
]
Leonard Xu commented on FLINK-38235:
------------------------------------
Yes, it's a limitation by design. The underlying reason is flink job
restoration can happen both in snapshot phase and binlog phase, it's hard to
distinguish the purpose of users
> Unable to reliably detect newly added tables with MySQL CDC
> -----------------------------------------------------------
>
> Key: FLINK-38235
> URL: https://issues.apache.org/jira/browse/FLINK-38235
> Project: Flink
> Issue Type: Bug
> Components: Flink CDC
> Affects Versions: cdc-3.3.0
> Reporter: Max Feng
> Priority: Major
>
> It appears to not be possible to configure the MySQL CDC 3.3.0 source such
> that newly added tables will be read regardless of their creation time. It
> appears that:
> scan.newly-added-table.enabled=true -> doesn't capture new tables in binlog
> mode
> scan.binlog.newly-added-table.enabled=true -> doesn't snapshot new tables
> both true -> Exception in the source factory
> Our goal is that:
> * When resuming from state, newly added tables are snapshot
> * When in binlog reading mode, newly created tables are captured
> Is this a limitation, or is there a configuration that works?
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
