[
https://issues.apache.org/jira/browse/FLINK-17012?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17084135#comment-17084135
]
Wenlong Lyu edited comment on FLINK-17012 at 4/15/20, 2:37 PM:
---------------------------------------------------------------
Treating initialization as part of deploying is also a reasonable idea.
How about we continue FLINK-4714: add an `initialize()` in AbstractInvokable
and call initialize in task before task transfer to running? Currently there
are similar functionalities in both StreamTask and BatchTask:
StreamTask#beforeInvoke and BatchTask#initialize which is called in invoke(),
the different is that; there are some common initialization in BatchTask#invoke
instead of BatchTask#initialize. we may need a bit more refactor for BatchTask
was (Author: wenlong.lwl):
Treating initialization as part of deploying is also a reasonable idea.
How about we continue FLINK-1474: add an `initialize()` in AbstractInvokable
and call initialize in task before task transfer to running? Currently there
are similar functionalities in both StreamTask and BatchTask:
StreamTask#beforeInvoke and BatchTask#initialize which is called in invoke(),
the different is that; there are some common initialization in BatchTask#invoke
instead of BatchTask#initialize. we may need a bit more refactor for BatchTask
> Expose stage of task initialization
> -----------------------------------
>
> Key: FLINK-17012
> URL: https://issues.apache.org/jira/browse/FLINK-17012
> Project: Flink
> Issue Type: Improvement
> Components: Runtime / Metrics, Runtime / Task
> Reporter: Wenlong Lyu
> Priority: Major
>
> Currently a task switches to running before fully initialized, does not take
> state initialization and operator initialization(#open ) in to account, which
> may take long time to finish. As a result, there would be a weird phenomenon
> that all tasks are running but throughput is 0.
> I think it could be good if we can expose the initialization stage of tasks.
> What to you think?
--
This message was sent by Atlassian Jira
(v8.3.4#803005)
