[
https://issues.apache.org/jira/browse/FLINK-4125?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15394020#comment-15394020
]
Maximilian Michels commented on FLINK-4125:
-------------------------------------------
Hi! This has already been fixed for the upcoming release. I overlooked your
JIRA issue when I opened FLINK-4139.
> Yarn CLI incorrectly calculates slotsPerTM when parallelism < task manager
> count
> --------------------------------------------------------------------------------
>
> Key: FLINK-4125
> URL: https://issues.apache.org/jira/browse/FLINK-4125
> Project: Flink
> Issue Type: Bug
> Reporter: Chris Hogue
> Priority: Minor
>
> The FlinkYarnSessionCli class tries to calculate how many slots each
> taskmanager gets based on the user-supplied parallelism option. It's using an
> int, so when -p < -yn it is truncated to 0 and the cluster then fails to
> start, saying slots per TM must be positive. And example command to trigger
> this would be:
> bin/flink run -m yarn-cluster -d -yn 20 -ys 2 -p 5 <a jar>
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
