You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@flink.apache.org by "Jiayi Liao (JIRA)" <ji...@apache.org> on 2018/10/23 12:31:00 UTC

[jira] [Commented] (FLINK-10654) AllowNonRestoredState should skip state that cannot be restored

    [ https://issues.apache.org/jira/browse/FLINK-10654?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16660567#comment-16660567 ] 

Jiayi Liao commented on FLINK-10654:
------------------------------------

From my perspective, the state is restore in the initialization in the stream tasks, which means that we can add AllowNonRestoredState as an option in stream task's configuration. However, the problem is how to notify other tasks that the restore of nth operator fails, if we want to make it consistent in all tasks?

> AllowNonRestoredState should skip state that cannot be restored
> ---------------------------------------------------------------
>
>                 Key: FLINK-10654
>                 URL: https://issues.apache.org/jira/browse/FLINK-10654
>             Project: Flink
>          Issue Type: Wish
>            Reporter: Jiayi Liao
>            Assignee: Jiayi Liao
>            Priority: Major
>
> We have a -n/--allowNonRestoredState option when we submit a job from command line. I know that it's used to ignore new operators' state, but I think it will be better if we can support that it can ignore those operator's state which throws exceptions during restore.
> In this way, we can still recover most states in other operators after we modify only a few source codes that are related to states.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)