[ 
https://issues.apache.org/jira/browse/TEZ-2278?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
 ]

Jason Lowe updated TEZ-2278:
----------------------------
    Attachment: TEZ-2278.001.patch

This technically could be solved with the data already being posted in the 
history, but the UI would have to fetch every attempt for each task and 
determine the earliest start time of the attempts to calculate the start time 
of the task.  That sounds like an expensive solution.

The task started event has the information we need as it has both a scheduled 
and started timestamp.  However that event is sent before any attempt is 
launched, so scheduled time == launched time in light of no known launch time.  
We could delay sending the tasks started event until after an attempt is 
launched, but then the recovery and history streams would have attempt start 
events occurring _before_ the task start events which would probably confuse 
current consumers of those streams.

Here's a patch that has the task emit another task started event once it 
transitions from SCHEDULED to RUNNING.  This event is _not_ recorded in the 
recovery stream to avoid confusion in the recovery parsing.


> Tez UI start/end time and duration shown are wrong for tasks
> ------------------------------------------------------------
>
>                 Key: TEZ-2278
>                 URL: https://issues.apache.org/jira/browse/TEZ-2278
>             Project: Apache Tez
>          Issue Type: Bug
>          Components: UI
>    Affects Versions: 0.6.0
>            Reporter: Rohini Palaniswamy
>            Assignee: Jonathan Eagles
>         Attachments: TEZ-2278.001.patch, screenshot-1.png, screenshot-2.png, 
> screenshot-3.png, screenshot-4.png
>
>
>  Observing lot of time discrepancies between vertex, task and swinlane views. 



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

Reply via email to