Naresh P R created TEZ-4443:
-------------------------------
Summary: Provide Tez AM/Task container range instead of fixed size
task containers (Metric based AM/task re-attempt with increased container size)
Key: TEZ-4443
URL: https://issues.apache.org/jira/browse/TEZ-4443
Project: Apache Tez
Issue Type: New Feature
Reporter: Naresh P R
Currently Tez supports only fixed size AM/Task container per execution.
* We assume task OOME as fatal & not re-attempt but fail the DAG. Instead if
we can get min/max container range and based on current executed task metrics,
Tez AM should be able to re-attempt same task with higher container size till
the maxRange / max re-attempts get exhausted.
* Similarly incase of AM OOME, can we utilize existing execution metrics to
re-attempt the same DAG with increased AM container
--
This message was sent by Atlassian Jira
(v8.20.10#820010)