[
https://issues.apache.org/jira/browse/TEZ-2929?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Sreenath Somarajapuram updated TEZ-2929:
----------------------------------------
Attachment: TEZ-2929.1.patch
[~hitesh]
Please let me know if the following approach is ok.
Its created considering tez.yarn.ats.event.flush.timeout.millis.
1. Load data from ATS
2. If dag/any vertex is running, start polling from AM
3 -- Polling ---
3.a If AM returns, update UI based on polled data
3.b.If AM is not reachable, ask RM for app status.
3.c If app.isComplete reload from ATS
3.d If not app.isComplete, show errorbar and continue polling
3.e If RM is not reachable, show errorbar and stop polling
----------------
4. On reload from ATS, If dag/any vertex is running, start polling with a start
delay of 10 sec
Pros:
- UI will update irrespective of the latency
Cons:
- If the application was killed, and application history is not available in
ATS, the DAG details page (only page) will continue polling
3 -> 3.b -> 3.c -> 4, continue to loop
> Tez UI: Dag details page displays vertices to be running even when dag have
> completed.
> --------------------------------------------------------------------------------------
>
> Key: TEZ-2929
> URL: https://issues.apache.org/jira/browse/TEZ-2929
> Project: Apache Tez
> Issue Type: Bug
> Reporter: Sreenath Somarajapuram
> Assignee: Sreenath Somarajapuram
> Attachments: TEZ-2929.1.patch
>
>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
