[
https://issues.apache.org/jira/browse/FLINK-31510?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17702427#comment-17702427
]
Shilun Fan commented on FLINK-31510:
------------------------------------
[~lemonjing] we use int32 for memory now, if a cluster has 10k nodes, each node
has 210G memory, we will get a negative total cluster memory.
> Use getMemorySize instead of getMemory
> --------------------------------------
>
> Key: FLINK-31510
> URL: https://issues.apache.org/jira/browse/FLINK-31510
> Project: Flink
> Issue Type: Improvement
> Components: Deployment / YARN
> Affects Versions: 1.18.0
> Reporter: Shilun Fan
> Priority: Major
> Labels: pull-request-available
>
> In YARN-4844, use getMemorySize instead of getMemory, because using int to
> represent memory may exceed the bounds in some cases and produce negative
> numbers.
> This change was merged in HADOOP-2.8.0, we should use getMemorySize instead
> of getMemory.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
