Merge pull request kubevirt#886 from davidvossel/fix_881

Add functional test to verify pod termination

cmd/registry-disk-v1alpha/entry-point.sh

@@ -19,6 +19,8 @@
 
 # https://fedoraproject.org/wiki/Scsi-target-utils_Quickstart_Guide
 
+trap 'echo "Graceful exit"; exit 0' SIGINT SIGQUIT SIGTERM
+
 if [ -z "$COPY_PATH" ]; then
 	echo "COPY_PATH variable not set"
 	exit 1

cmd/virt-launcher/entrypoint.sh

@@ -1,6 +1,13 @@
 #!/bin/bash
 set +e
 
+_term() { 
+  echo "caught signal"
+  kill -TERM "$virt_launcher_pid" 2>/dev/null
+}
+
+trap _term SIGTERM SIGINT SIGQUIT
+
 # HACK
 # Try to create /dev/kvm if not present
 if [ ! -e /dev/kvm ]; then
@@ -10,13 +17,24 @@ fi
 chown :qemu /dev/kvm
 chmod 660 /dev/kvm
 
-
 # Cockpit/OCP hack to all shoing the vm terminal
 mv /usr/bin/sh /usr/bin/sh.orig
 mv /sh.sh /usr/bin/sh
 chmod +x /usr/bin/sh
 
-./virt-launcher $@
+./virt-launcher $@ &
+virt_launcher_pid=$!
+while true; do
+	if ! [ -d /proc/$virt_launcher_pid ]; then
+		break;
+	fi
+	sleep 1
+done
+# call wait after we know the pid has exited in order
+# to get the return code. If we call wait before the pid
+# exits, wait will actually return early when we forward
+# the trapped signal in _trap(). We don't want that.
+wait -n $virt_launcher_pid
 rc=$?
 
 echo "virt-launcher exited with code $rc"

pkg/virt-launcher/monitor.go

@@ -179,6 +179,16 @@ func NewProcessMonitor(commandPrefix string,
 	}
 }
 
+func (mon *monitor) isGracePeriodExpired() bool {
+	if mon.gracePeriodStartTime != 0 {
+		now := time.Now().UTC().Unix()
+		if (now - mon.gracePeriodStartTime) > int64(mon.gracePeriod) {
+			return true
+		}
+	}
+	return false
+}
+
 func (mon *monitor) refresh() {
 	if mon.isDone {
 		log.Log.Error("Called refresh after done!")
@@ -187,6 +197,8 @@ func (mon *monitor) refresh() {
 
 	log.Log.Debugf("Refreshing. CommandPrefix %s pid %d", mon.commandPrefix, mon.pid)
 
+	expired := mon.isGracePeriodExpired()
+
 	// is the process there?
 	if mon.pid == 0 {
 		var err error
@@ -200,6 +212,9 @@ func (mon *monitor) refresh() {
 			if mon.timeout > 0 && elapsed >= mon.timeout {
 				log.Log.Infof("%s not found after timeout", mon.commandPrefix)
 				mon.isDone = true
+			} else if expired {
+				log.Log.Infof("%s not found after grace period expired", mon.commandPrefix)
+				mon.isDone = true
 			}
 			return
 		}
@@ -219,12 +234,9 @@ func (mon *monitor) refresh() {
 		return
 	}
 
-	if mon.gracePeriodStartTime != 0 {
-		now := time.Now().UTC().Unix()
-		if (now - mon.gracePeriodStartTime) > int64(mon.gracePeriod) {
-			log.Log.Infof("Grace Period expired, shutting down.")
-			mon.shutdownCallback(mon.pid)
-		}
+	if expired {
+		log.Log.Infof("Grace Period expired, shutting down.")
+		mon.shutdownCallback(mon.pid)
 	}
 
 	return
@@ -252,9 +264,6 @@ func (mon *monitor) monitorLoop(startTimeout time.Duration, signalChan chan os.S
 			mon.refresh()
 		case s := <-signalChan:
 			log.Log.Infof("Received signal %d.", s)
-			if mon.pid == 0 {
-				continue
-			}
 
 			if mon.gracePeriodStartTime != 0 {
 				continue

pkg/virt-launcher/monitor_test.go

@@ -145,6 +145,30 @@ var _ = Describe("VirtLauncher", func() {
 				Expect(exited).To(Equal(true))
 			})
 
+			It("verify monitor loop exits when signal arrives and no pid is present", func() {
+				signalChannel := make(chan os.Signal, 1)
+				done := make(chan string)
+
+				go func() {
+					mon.monitorLoop(1*time.Second, signalChannel)
+					done <- "exit"
+				}()
+
+				time.Sleep(time.Second)
+
+				signalChannel <- syscall.SIGQUIT
+				noExitCheck := time.After(5 * time.Second)
+				exited := false
+
+				select {
+				case <-noExitCheck:
+				case <-done:
+					exited = true
+				}
+
+				Expect(exited).To(Equal(true))
+			})
+
 			It("verify graceful shutdown trigger works", func() {
 				signalChannel := make(chan os.Signal, 1)
 				done := make(chan string)

tests/vmlifecycle_test.go

@@ -37,6 +37,7 @@ import (
 
 	"kubevirt.io/kubevirt/pkg/api/v1"
 	"kubevirt.io/kubevirt/pkg/kubecli"
+	"kubevirt.io/kubevirt/pkg/virt-controller/services"
 	"kubevirt.io/kubevirt/tests"
 )
 
@@ -351,6 +352,32 @@ var _ = Describe("Vmlifecycle", func() {
 	})
 
 	Describe("Delete a VM", func() {
+		Context("with an active pod.", func() {
+			It("should result in pod being terminated", func(done Done) {
+
+				By("Creating the VM")
+				obj, err := virtClient.VM(tests.NamespaceTestDefault).Create(vm)
+				Expect(err).ToNot(HaveOccurred())
+				tests.WaitForSuccessfulVMStart(obj)
+
+				By("Verifying VM's pod is active")
+				pods, err := virtClient.CoreV1().Pods(tests.NamespaceTestDefault).List(services.UnfinishedVMPodSelector(vm))
+				Expect(err).ToNot(HaveOccurred())
+				Expect(len(pods.Items)).To(Equal(1))
+
+				By("Deleting the VM")
+				Expect(virtClient.VM(vm.Namespace).Delete(obj.Name, &metav1.DeleteOptions{})).To(Succeed())
+
+				By("Verifying VM's pod terminates")
+				Eventually(func() int {
+					pods, err := virtClient.CoreV1().Pods(tests.NamespaceTestDefault).List(services.UnfinishedVMPodSelector(vm))
+					Expect(err).ToNot(HaveOccurred())
+					return len(pods.Items)
+				}, 75, 0.5).Should(Equal(0))
+
+				close(done)
+			}, 90)
+		})
 		Context("with grace period greater than 0", func() {
 			It("should run graceful shutdown", func(done Done) {
 				nodes, err := virtClient.CoreV1().Nodes().List(metav1.ListOptions{})