Skip to content

[SPARK-58725][K8S] Resolve the pod before deciding whether --kill found it #57948

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
LuciferYang wants to merge 2 commits into
base: master
Choose a base branch
from
+27 −2
Open

[SPARK-58725][K8S] Resolve the pod before deciding whether --kill found it #57948

Show file tree
Hide file tree
Changes from all commits
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
c198a91
[MINOR][K8S] Resolve the pod before deciding whether --kill found it
LuciferYang Aug 11, 2026
2983bdd
[SPARK-58725][K8S][FOLLOWUP] Add the JIRA id and drop an unreachable …
LuciferYang Aug 12, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
6 changes: 5 additions & 1 deletion ...gers/kubernetes/core/src/main/scala/org/apache/spark/deploy/k8s/submit/K8sSubmitOps.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -50,7 +50,11 @@ private class KillApplication extends K8sSubmitOp {
(implicit client: KubernetesClient): Unit = {
val podToDelete = getPod(namespace, pName)

if (Option(podToDelete).isDefined) {
// `getPod` returns a request handle, which is never null; only resolving it reports whether
// the pod exists. Without the `get()` the check below is always true, so a name that is not
// in the cluster would issue a delete that the API server answers with a swallowed 404 and
// report nothing to the user. `ListStatus.executeOnPod` resolves it the same way.
if (Option(podToDelete.get()).isDefined) {
getGracePeriod(sparkConf) match {
case Some(period) => podToDelete.withGracePeriod(period).delete()
case _ => podToDelete.delete()
Expand Down
23 changes: 22 additions & 1 deletion .../kubernetes/core/src/test/scala/org/apache/spark/deploy/k8s/submit/K8sSubmitOpSuite.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -25,7 +25,7 @@ import io.fabric8.kubernetes.api.model._
import io.fabric8.kubernetes.client.{KubernetesClient, PropagationPolicyConfigurable}
import io.fabric8.kubernetes.client.dsl.{Deletable, NamespaceListVisitFromServerGetDeleteRecreateWaitApplicable, PodResource}
import org.mockito.{ArgumentMatchers, Mock, MockitoAnnotations}
import org.mockito.Mockito.{times, verify, when}
import org.mockito.Mockito.{never, times, verify, when}
import org.scalatest.BeforeAndAfter

import org.apache.spark.{SparkConf, SparkFunSuite}
Expand All @@ -38,6 +38,7 @@ import org.apache.spark.scheduler.cluster.k8s.ExecutorLifecycleTestUtils.TEST_SP
class K8sSubmitOpSuite extends SparkFunSuite with BeforeAndAfter {
private val driverPodName1 = "driver1"
private val driverPodName2 = "driver2"
private val missingPodName = "driver3"
private val driverPod1 = buildDriverPod(driverPodName1, "1")
private val driverPod2 = buildDriverPod(driverPodName2, "2")
private val podList = List(driverPod1, driverPod2)
Expand All @@ -55,6 +56,13 @@ class K8sSubmitOpSuite extends SparkFunSuite with BeforeAndAfter {
@Mock
private var driverPodOperations2: PodResource = _

// A handle for a pod that does not exist. `withName` returns a handle regardless, and `get`
// is what reports absence by returning null, so the missing pod has to be a real mock rather
// than an unstubbed name: an unstubbed `withName` would return null and make the buggy
// `Option(handle).isDefined` check take the false branch for the wrong reason.
@Mock
private var missingPodOperations: PodResource = _

@Mock
private var kubernetesClient: KubernetesClient = _

Expand All @@ -76,10 +84,12 @@ class K8sSubmitOpSuite extends SparkFunSuite with BeforeAndAfter {
when(podOperations.inNamespace(namespace)).thenReturn(podsWithNamespace)
when(podsWithNamespace.withName(driverPodName1)).thenReturn(driverPodOperations1)
when(podsWithNamespace.withName(driverPodName2)).thenReturn(driverPodOperations2)
when(podsWithNamespace.withName(missingPodName)).thenReturn(missingPodOperations)
when(driverPodOperations1.get).thenReturn(driverPod1)
when(driverPodOperations1.delete()).thenReturn(Arrays.asList(new StatusDetails))
when(driverPodOperations2.get).thenReturn(driverPod2)
when(driverPodOperations2.delete()).thenReturn(Arrays.asList(new StatusDetails))
doReturn(null).when(missingPodOperations).get
}

test("List app status") {
Expand Down Expand Up @@ -120,6 +130,17 @@ class K8sSubmitOpSuite extends SparkFunSuite with BeforeAndAfter {
verify(deletable, times(1)).delete()
}

test("SPARK-58725: Kill app that does not exist") {
implicit val kubeClient: KubernetesClient = kubernetesClient
val killApp = new KillApplication
killApp.printStream = err
killApp.executeOnPod(missingPodName, Option(namespace), new SparkConf())
// scalastyle:off
verify(err).println(ArgumentMatchers.eq("Application not found."))
// scalastyle:on
verify(missingPodOperations, never()).delete()
}

test("Kill multiple apps with glob without gracePeriod") {
implicit val kubeClient: KubernetesClient = kubernetesClient
val killApp = new KillApplication
Expand Down