[
https://issues.apache.org/jira/browse/FLINK-33051?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17765162#comment-17765162
]
Matt Wang commented on FLINK-33051:
-----------------------------------
[~pgaref] Let me think about it. I don't think it's entirely reasonable to
replace GlobalFailureHandler with LabeledGlobalFailureHandler. Now the
implementation class of GlobalFailureHandler is mainly in the scheduler, while
the implementation class of LabeledGlobalFailureHandler is
`org.apache.flink.runtime.scheduler.adaptive.State`. The `failureLabels`
information is obtained in the scheduler (DefaultScheduler/AdaptiveScheduler),
and the API of Scheduler's `handleGlobalFailure(Throwable cause,
CompletableFuture<Map<String, String>> failureLabels)` is not reasonable.
However, GlobalFailureHandler and LabeledGlobalFailureHandler have overlapping
functionality. Here, I tend to unify GlobalFailureHandler and
LabeledGlobalFailureHandler into one interface, and State and scheduler will go
through different methods. I will submit an MR today, and if you have time,
help me review it.
> GlobalFailureHandler interface should be retired in favor of
> LabeledGlobalFailureHandler
> ----------------------------------------------------------------------------------------
>
> Key: FLINK-33051
> URL: https://issues.apache.org/jira/browse/FLINK-33051
> Project: Flink
> Issue Type: Sub-task
> Components: Runtime / Coordination
> Reporter: Panagiotis Garefalakis
> Assignee: Matt Wang
> Priority: Minor
>
> FLIP-304 introduced `LabeledGlobalFailureHandler` interface that is an
> extension of `GlobalFailureHandler` interface. The later can thus be removed
> in the future to avoid the existence of interfaces with duplicate functions.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
