zwangsheng commented on code in PR #40118:
URL: https://github.com/apache/spark/pull/40118#discussion_r1151449690
##########
resource-managers/kubernetes/core/src/main/scala/org/apache/spark/deploy/k8s/submit/LoggingPodStatusWatcher.scala:
##########
@@ -95,6 +100,27 @@ private[k8s] class LoggingPodStatusWatcherImpl(conf:
KubernetesDriverConf)
this.notifyAll()
}
+ override def getDriverExitCode(): Option[Int] = synchronized {
+ if (hasCompleted()) {
+ val driverContainerName =
conf.get(KUBERNETES_DRIVER_PODTEMPLATE_CONTAINER_NAME)
+ .getOrElse(DEFAULT_DRIVER_CONTAINER_NAME)
+ pod.foreach { p =>
+ try {
+ return Some(p.getStatus.getContainerStatuses.asScala
+ .filter(driverContainerName == _.getName)
+ .head.getState.getTerminated.getExitCode)
+ } catch {
+ case _: NullPointerException =>
+ logError("Fail to find completed driver container exit code")
+ return None
+ }
+ }
+ throw new SparkException("Fail to get driver exit code, when the
application completed")
+ } else {
+ throw new SparkException("Call getDriverExitCode() when the application
has not completed")
Review Comment:
Thank you for looking at the code in detail.
Four scenarios are considered here:
* Case 1: which normal won't be triggered under normal case, placed here to
prevent the function from being misused.
* Case 2: Successful case
* Case 3: Only trigger if the app has completed but pod without exit code
* Case 4: due to we didn't get the specified driver container(Neither
`KUBERNETES_DRIVER_PODTEMPLATE_CONTAINER_NAME` nor
`DEFAULT_DRIVER_CONTAINER_NAME `)
> PS. In `KubernetesUtils.loadPodFromTemplate` function logic, we read the
first container from user's template file, not hardwired to
`KUBERNETES_DRIVER_PODTEMPLATE_CONTAINER_NAME` or
`DEFAULT_DRIVER_CONTAINER_NAME`.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
