zlzhang0122 created FLINK-26129:
-----------------------------------
Summary: Using fail method instead of kill method when yarn
application deployment error occures
Key: FLINK-26129
URL: https://issues.apache.org/jira/browse/FLINK-26129
Project: Flink
Issue Type: Improvement
Components: Deployment / YARN
Affects Versions: 1.14.3
Reporter: zlzhang0122
Fix For: 1.15.0
When deploy an application using yarn, it add a DeploymentFailureHook and if
the deployment duration exceed the applicationSubmitTimeout(this situation
always occures when the resource is not sufficient,etc.).
Now it uses yarnClient.killApplication() to kill fail this deployment, I think
maybe we should using yarnClient.failApplicationAttempt() to fail the
deployment. IMO, using fail method instead of kill method can trigger the retry
action and is more reasonable.
The change is very simple and only affect the yarn application deployment.
--
This message was sent by Atlassian Jira
(v8.20.1#820001)
