[
https://issues.apache.org/jira/browse/TEZ-1142?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14004129#comment-14004129
]
Bikas Saha commented on TEZ-1142:
---------------------------------
Not quite. There may be other cases where early exit is possible via user logic
and the DAG may pass. This jira is about failing the dag asap. Such dags are
very likely to fail after the existing 4 attempts failed condition gets hit.
This jira wants to speed that up by creating another failure policy in addition
to the 4 attempts == failure policy.
> Bail out early if a vertex has too many failures
> ------------------------------------------------
>
> Key: TEZ-1142
> URL: https://issues.apache.org/jira/browse/TEZ-1142
> Project: Apache Tez
> Issue Type: Sub-task
> Reporter: Bikas Saha
>
> If there are a high number of failures in a vertex then bail out early
> instead of waiting for 4 failures of the same task.
> Lets say the vertex sees N consecutive failures without any successful task
> completion. Thats probably good enough evidence to infer that there is some
> bug in the code for the tasks in that vertex. Bailing out early wastes less
> resources.
--
This message was sent by Atlassian JIRA
(v6.2#6252)
