[ 
https://issues.apache.org/jira/browse/FLINK-13848?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16919138#comment-16919138
 ] 

Biao Liu commented on FLINK-13848:
----------------------------------

[~till.rohrmann]

Sorry I didn't describe it clearly.

In this refactoring, the {{triggerCheckpoint}} would be separated into several 
stages. For example A -> B -> C. A is "start triggering", running in main 
thread. B is doing some heavy initializations, running in IO threads. C is 
sending the trigger message to tasks, running in main thread. The workflow 
should be like:
 # First A is triggered and executed in main thread. A launches the B into IO 
threads, and returns immediately.
 # B is executed in IO thread. When it's finished, a callback of triggering C 
executes (maybe through {{CompletableFuture#thenRunAsync}} with main thread 
executor).
 # C is executed in main thread. After then the next A is scheduled after a 
delay.

It's A -> B -> C -> A ... There would be no competition. That's what we really 
want.

But for the {{scheduleAtFixedDelay}} way, what we scheduled here is just A. The 
next A would be scheduled when the prior A is finished.
 It's A -> B -> C
         \
           -> A -> B -> C

So I think we should abandon this ticket. Or convert it to a normal issue not a 
subtask of this refactoring. What do you think?

> Support “scheduleAtFixedRate/scheduleAtFixedDelay” in 
> RpcEndpoint#MainThreadExecutor
> ------------------------------------------------------------------------------------
>
>                 Key: FLINK-13848
>                 URL: https://issues.apache.org/jira/browse/FLINK-13848
>             Project: Flink
>          Issue Type: Sub-task
>          Components: Runtime / Coordination
>            Reporter: Biao Liu
>            Priority: Major
>             Fix For: 1.10.0
>
>
> Currently the methods “scheduleAtFixedRate/scheduleAtFixedDelay" of 
> {{RpcEndpoint#MainThreadExecutor}} are not implemented. Because there was no 
> requirement on them before.
> Now we are planning to implement these methods to support periodic checkpoint 
> triggering.



--
This message was sent by Atlassian Jira
(v8.3.2#803003)

Reply via email to