You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@tez.apache.org by "Bikas Saha (JIRA)" <ji...@apache.org> on 2015/04/03 01:28:54 UTC

[jira] [Commented] (TEZ-2268) Add an event to indicate that all events are available

    [ https://issues.apache.org/jira/browse/TEZ-2268?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14393721#comment-14393721 ] 

Bikas Saha commented on TEZ-2268:
---------------------------------

Can you please describe the problem/scenario for this? 
Is this for all events or only data movement events? Does this mean globally all events for an input are available. Or is it that a particular task's outputs are available for an input?
E.g. depending on the scenario, perhaps, adding a booleas isLast flag on the DataMovement enough should be enough to piggyback the required information instead of using another (and new) event.


> Add an event to indicate that all events are available
> ------------------------------------------------------
>
>                 Key: TEZ-2268
>                 URL: https://issues.apache.org/jira/browse/TEZ-2268
>             Project: Apache Tez
>          Issue Type: Improvement
>            Reporter: Siddharth Seth
>            Assignee: Siddharth Seth
>
> Instead of relying on event counts, it'll be useful to have a single event indicate that all events have been generated.
> We'll have to guarantee ordering of events to be after all events are actually sent over.
> EdgeManagers, unless specialized, cannot generate these events. They could, however, be linked to source Vertex completion.
> Inputs / Outputs could use this to simplify tracking. Failures are the only events which could come in after this - which can be ignored by Inputs if they have already fetched the relevant data.



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