[
https://issues.apache.org/jira/browse/FLINK-10996?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16706685#comment-16706685
]
aitozi commented on FLINK-10996:
--------------------------------
Hi, [~StephanEwen]
I think it can be resolved by supporting the api to set the related expire ttl
time in flink cep. If it is set, then cep operator can init the state with ttl
when creating the nfa/sharedbuffer related state. I will work on it this week.
> Enable state ttl in cep
> -----------------------
>
> Key: FLINK-10996
> URL: https://issues.apache.org/jira/browse/FLINK-10996
> Project: Flink
> Issue Type: Improvement
> Components: CEP
> Affects Versions: 1.7.1
> Reporter: aitozi
> Assignee: aitozi
> Priority: Major
> Fix For: 1.7.1
>
>
> In flink cep, although the within clause can clean up the data when data is
> expire. But If the specific key data never come again, the partial matched
> data will not be deleted again which will lead to the state leak. With the
> state ttl feature support by flink runtime, we has the ability to expire the
> correlative data asyncly .
--
This message was sent by Atlassian JIRA
(v7.6.3#76005)
