[
https://issues.apache.org/jira/browse/FLINK-34792?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17892701#comment-17892701
]
wangli commented on FLINK-34792:
--------------------------------
I think there are two problems here.
# If checkpointing is not enabled, assignerStatus will always remain
INITIAL_ASSIGNING, and the MySQL source will not start reading incremental data.
# Even if checkpoint is enabled, if the interval is set to a large value,
similar problems will occur. !image-2024-10-25-16-00-58-373.png!
> [Bug] Incremental source never switches to streaming mode with disabled
> checkpointing and parallelism > 1
> ---------------------------------------------------------------------------------------------------------
>
> Key: FLINK-34792
> URL: https://issues.apache.org/jira/browse/FLINK-34792
> Project: Flink
> Issue Type: Bug
> Components: Flink CDC
> Reporter: Flink CDC Issue Import
> Priority: Major
> Labels: github-import
> Attachments: image-2024-10-25-16-00-58-373.png
>
>
> ### Search before asking
> - [X] I searched in the
> [issues|https://github.com/ververica/flink-cdc-connectors/issues) and found
> nothing similar.
> ### Flink version
> 1.13.6, 1.14.4, 1.15.2, 1.16.0, 1.17.0
> ### Flink CDC version
> 2.3.0, 2.4-SNAPSHOT
> ### Database and its version
> Any connector with incremental source implementation
> ### Minimal reproduce step
> For instance, remove
> ```sql
> SET 'execution.checkpointing.interval' = '3s';
> ```
> from `MySqlE2eITCase#testMySqlCDC(]`
> ### What did you expect to see?
> Failed job or at least error message describing that the user must enable
> checkpointing for the connector to work correctly
> ### What did you see instead?
> If parallelism of incremental source is greater than one, implementations of
> `SnapshotSplitAssigner` wait for completed checkpoint event to set flag
> field, that triggers `HybridSplitAssigner` to switch to streaming mode. If
> checkpointing is disabled, then flag field is never set to it's final state,
> which leads to the fact that the source is forever stuck in the snapshot
> mode. So, I think we need to forbid users to run jobs with incremental
> sources without enabled checkpointing.
> ### Anything else?
> _No response_
> ### Are you willing to submit a PR?
> - [ ] I'm willing to submit a PR!
> ---------------- Imported from GitHub ----------------
> Url: https://github.com/apache/flink-cdc/issues/2165
> Created by: [tigrulya-exe|https://github.com/tigrulya-exe]
> Labels: bug,
> Created at: Tue May 30 15:39:27 CST 2023
> State: open
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
