[
https://issues.apache.org/jira/browse/TEZ-3630?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15879901#comment-15879901
]
Siddharth Seth commented on TEZ-3630:
-------------------------------------
+1. Thanks [~Sreenath]
Discussed offline about separate isComplete implementation for DAG/App. Looks
like the same is also used for Vertices, Tasks and TaskAttempts. This happens
to work since most states are shared. One missing from YARN is ENDED. Think we
need another jira to separate out these states for the different entities.
> Tez UI: Use DAG status for controlling auto-refresh polling.
> ------------------------------------------------------------
>
> Key: TEZ-3630
> URL: https://issues.apache.org/jira/browse/TEZ-3630
> Project: Apache Tez
> Issue Type: Bug
> Reporter: Sreenath Somarajapuram
> Assignee: Sreenath Somarajapuram
> Attachments: TEZ-3630.1.patch
>
>
> Issue: UI throws error on trying to access details of a SUCCEEDED DAG in a
> RUNNING app.
> Cause: Right now we depend on app status to control polling, and its not
> mandatory for the AM to have details about non-running DAGs run under it.
> Hence error is thrown on polling AM for old DAG details.
> Note:
> - Current implementation uses app status as it considers a possible case
> where DAG could move to SUCCEEDED state and then go back to RUNNING.
> - It also presumed the AM to have data about all DAGs run under it.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
