[
https://issues.apache.org/jira/browse/TWILL-230?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=15954382#comment-15954382
]
ASF GitHub Bot commented on TWILL-230:
--------------------------------------
Github user chtyim commented on a diff in the pull request:
https://github.com/apache/twill/pull/53#discussion_r109553180
--- Diff:
twill-yarn/src/main/java/org/apache/twill/yarn/YarnTwillController.java ---
@@ -322,7 +314,46 @@ private boolean hasRun(YarnApplicationState state) {
@Override
public ResourceReport getResourceReport() {
- // in case the user calls this before starting, return null
+ // Only has resource report if the app is running.
+ if (state() != State.RUNNING) {
+ return null;
+ }
+ ResourceReportClient resourcesClient = getResourcesClient();
return (resourcesClient == null) ? null : resourcesClient.get();
}
+
+ /**
+ * Returns the {@link ResourceReportClient} for fetching resource report
from the AM.
+ * It first consults the RM for the tracking URL and get the resource
report from there.
+ */
+ @Nullable
+ private ResourceReportClient getResourcesClient() {
+ YarnApplicationReport report = processController.getReport();
+ List<URL> urls = new ArrayList<>(2);
--- End diff --
Because we store the tracking URL and the original tracking URL.
> Retrieval of ResourceReport via TwillController may fail
> --------------------------------------------------------
>
> Key: TWILL-230
> URL: https://issues.apache.org/jira/browse/TWILL-230
> Project: Apache Twill
> Issue Type: Bug
> Reporter: Terence Yim
> Fix For: 0.11.0
>
>
> When starting a new {{YarnTwillRunnerService}}, it discovers all existing
> applications and get the tracker URL from the application report. However,
> the operation might fail if the application was started with a different user
> than the one who starts the {{YarnTwillRunnerService}} in a Kerberos cluster,
> resulting in failure to get resource report.
--
This message was sent by Atlassian JIRA
(v6.3.15#6346)
