[
https://issues.apache.org/jira/browse/YARN-7454?page=com.atlassian.jira.plugin.system.issuetabpanels:all-tabpanel
]
Jason Lowe updated YARN-7454:
-----------------------------
Description:
RMAppAttemptMetrics#getAggregateResourceUsage does a double-lookup on a
concurrent hash map, but the app could be removed from the map between the two
lookups:
{code}
RMApp rmApp = rmContext.getRMApps().get(attemptId.getApplicationId());
if (rmApp != null) {
RMAppAttempt currentAttempt =
rmContext.getRMApps().get(attemptId.getApplicationId()).getCurrentAppAttempt();
{code}
The attempt should be looked up within rmApp directly rather than redundantly
trying to retrieve the RMApp first.
> RMAppAttemptMetrics#getAggregateResourceUsage can NPE due to double lookup
> --------------------------------------------------------------------------
>
> Key: YARN-7454
> URL: https://issues.apache.org/jira/browse/YARN-7454
> Project: Hadoop YARN
> Issue Type: Bug
> Components: resourcemanager
> Reporter: Jason Lowe
> Priority: Minor
>
> RMAppAttemptMetrics#getAggregateResourceUsage does a double-lookup on a
> concurrent hash map, but the app could be removed from the map between the
> two lookups:
> {code}
> RMApp rmApp = rmContext.getRMApps().get(attemptId.getApplicationId());
> if (rmApp != null) {
> RMAppAttempt currentAttempt =
> rmContext.getRMApps().get(attemptId.getApplicationId()).getCurrentAppAttempt();
> {code}
> The attempt should be looked up within rmApp directly rather than redundantly
> trying to retrieve the RMApp first.
--
This message was sent by Atlassian JIRA
(v6.4.14#64029)
---------------------------------------------------------------------
To unsubscribe, e-mail: yarn-issues-unsubscr...@hadoop.apache.org
For additional commands, e-mail: yarn-issues-h...@hadoop.apache.org
