You are viewing a plain text version of this content. The canonical link for it is here.
Posted to dev@reef.apache.org by "Markus Weimer (JIRA)" <ji...@apache.org> on 2016/08/23 15:28:20 UTC

[jira] [Resolved] (REEF-1522) Optimize WaitForTaskRegistration

     [ https://issues.apache.org/jira/browse/REEF-1522?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel ]

Markus Weimer resolved REEF-1522.
---------------------------------
       Resolution: Fixed
    Fix Version/s: 0.16

Resolved via [#1092|https://github.com/apache/reef/pull/1092]

> Optimize WaitForTaskRegistration
> --------------------------------
>
>                 Key: REEF-1522
>                 URL: https://issues.apache.org/jira/browse/REEF-1522
>             Project: REEF
>          Issue Type: Improvement
>            Reporter: Julia
>            Assignee: Julia
>            Priority: Minor
>             Fix For: 0.16
>
>
> Currently in OperatorTopology WaitForTaskRegistration, we wait for parent first, then children. Each wait is in a retry/sleep loop until either found or timeout. After finding the first one, we will then look up for the second, and each loop is a blocking call.  
> We can improve it by looking up all of them in the same loop. In the end, if any of them cannot be found, we will throw exception. So that the entire look up is in one retry/sleep loop. 



--
This message was sent by Atlassian JIRA
(v6.3.4#6332)