[
https://issues.apache.org/jira/browse/YARN-3464?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=14484920#comment-14484920
]
zhihai xu commented on YARN-3464:
---------------------------------
This issue only happened for PRIVATE/APPLICATION resource Localization
We saw this issue happened when the PRIVATE LocalizerResourceRequestEvent
interleaved with PUBLIC LocalizerResourceRequestEvent in the following order:
PRIVATE1
PRIVATE2
..........
PRIVATEm
PUBLIC1
PUBLIC2
.....
PUBLICn
PRIVATEm+1
PRIVATEm+2
The last two PRIVATE LocalizerResourceRequestEvent is added after all previous
m PRIVATE LocalizerResourceRequestEvent are LOCALIZED due to the delay to
process n PUBLIC LocalizerResourceRequestEvent.
Then the container will stay at LOCALIZING state until it is killed by AM.
> Race condition in LocalizerRunner causes container localization timeout.
> ------------------------------------------------------------------------
>
> Key: YARN-3464
> URL: https://issues.apache.org/jira/browse/YARN-3464
> Project: Hadoop YARN
> Issue Type: Bug
> Components: nodemanager
> Reporter: zhihai xu
> Assignee: zhihai xu
> Priority: Critical
>
> Race condition in LocalizerRunner causes container localization timeout.
> Currently LocalizerRunner will kill the ContainerLocalizer when pending list
> for LocalizerResourceRequestEvent is empty.
> {code}
> } else if (pending.isEmpty()) {
> action = LocalizerAction.DIE;
> }
> {code}
> If a LocalizerResourceRequestEvent is added after LocalizerRunner kill the
> ContainerLocalizer due to empty pending list, this
> LocalizerResourceRequestEvent will never be handled.
> The container will stay at LOCALIZING state, until the container is killed by
> AM due to TASK_TIMEOUT.
--
This message was sent by Atlassian JIRA
(v6.3.4#6332)
