You are viewing a plain text version of this content. The canonical link for it is here.
Posted to issues@tez.apache.org by "Eric Wohlstadter (JIRA)" <ji...@apache.org> on 2019/01/07 21:52:00 UTC

[jira] [Commented] (TEZ-4028) Events not visible from proto history logging for s3a filesystem until dag completes.

    [ https://issues.apache.org/jira/browse/TEZ-4028?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=16736408#comment-16736408 ] 

Eric Wohlstadter commented on TEZ-4028:
---------------------------------------

[~harishjp]

I was initially confused why "_1" was being added.

Minor: Could you define that string as a constant with a descriptive name to indicate it is related to TEZ_HISTORY_LOGGING_PROTO_SPLIT_DAG_START?

> Events not visible from proto history logging for s3a filesystem until dag completes.
> -------------------------------------------------------------------------------------
>
>                 Key: TEZ-4028
>                 URL: https://issues.apache.org/jira/browse/TEZ-4028
>             Project: Apache Tez
>          Issue Type: Bug
>            Reporter: Harish Jaiprakash
>            Assignee: Harish Jaiprakash
>            Priority: Major
>         Attachments: TEZ-4028.01.patch
>
>
> The events are not visible in the files because  s3 filesystem
> * flush writes to local disk and only upload/commit to s3 on close.
> * does not support append
> As an initial fix we log the dag submitted, initialized and started events into a file and these can be read to get the dag plan, config from the AM. The counters are anyways not available until the dag completes.
> The in-progress information cannot be read, this can be obtained from the AM once we have the above events.



--
This message was sent by Atlassian JIRA
(v7.6.3#76005)