[
https://issues.apache.org/jira/browse/FLINK-34001?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17804208#comment-17804208
]
xuyang commented on FLINK-34001:
--------------------------------
When the data is queried, the cache will record it into the cache in the memory
through LRU. Therefore, if the cache miss occurs, it may put greater pressure
on the external database.
An effective method is using cache all, but the community has not yet
implemented it. For details, please refer to
[here|https://issues.apache.org/jira/browse/FLINK-17178]
If your job is with a CDC data source, I do not recommend that you use a state
TTL like 1ms to work around it, because it has no way to handle the correct
retraction message(-U, -D). Maybe you can handle it by using UDF or slightly
modifying the connector source code.
> Fix ambiguous document description towards the default value for configuring
> operator-level state TTL
> -----------------------------------------------------------------------------------------------------
>
> Key: FLINK-34001
> URL: https://issues.apache.org/jira/browse/FLINK-34001
> Project: Flink
> Issue Type: Improvement
> Components: Documentation, Table SQL / API
> Affects Versions: 1.18.0
> Reporter: yong yang
> Assignee: Jane Chan
> Priority: Major
> Labels: pull-request-available
>
> doc:
> https://nightlies.apache.org/flink/flink-docs-release-1.18/docs/dev/table/concepts/overview/#idle-state-retention-time
> The current TTL value for both left and right side is {{{}"0 ms"{}}}, which
> means the state retention is not enabled.
>
> but i test find :
> The current TTL value for both left and right side is {{{}"0 ms"{}}}, which
> means the state is permanence keep!
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
