Github user tgravescs commented on a diff in the pull request:
https://github.com/apache/spark/pull/1180#discussion_r15351507
--- Diff:
yarn/common/src/main/scala/org/apache/spark/scheduler/cluster/YarnClientSchedulerBackend.scala
---
@@ -115,7 +117,33 @@ private[spark] class YarnClientSchedulerBackend(
}
}
+ private def yarnApplicationStateCheckerThread(): Thread = {
+ var sparkContextStopped = false
+ val t = new Thread {
+ override def run() {
+ while (!sparkContextStopped) {
+ val report = client.getApplicationReport(appId)
+ val state = report.getYarnApplicationState()
+ if (state == YarnApplicationState.FINISHED || state ==
YarnApplicationState.KILLED
+ || state == YarnApplicationState.FAILED) {
+ logError(s"Yarn application already ended: $state")
+ sc.stop()
--- End diff --
so in testing this I saw this properly kick in (when I manually killed a
couple of the executors) and I saw stuff starting to shut down but then my
client just basically hung there.
last message on the screen was:
14/07/24 14:57:27 WARN BlockManagerMasterActor: Removing BlockManager
BlockManagerId(3, gsrd203n10.red.ygrid.yahoo.com, 40681, 0) with no recent
heart beats: 59378ms exceeds 45000ms
did you see this at all or did it close down properly for you?
This was testing on hadoop 0.23
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
