Migrate OCP-44493: configurable terminationGracePeriod for liveness and startup probes

Migrates test from openshift-tests-private to origin.

Test validates probe-level terminationGracePeriodSeconds for:
- Liveness probes with probe-level terminationGracePeriodSeconds (10s)
- Startup probes with probe-level terminationGracePeriodSeconds (10s)
- Liveness probes without probe-level (falls back to pod-level 60s)

The test creates pods with failing probes and verifies the time
difference between probe failure (Killing event) and container
restart (Started event) matches the expected termination grace
period within acceptable range.

Event matching logic parses 'oc describe pod' output for:
- Killing events with container name
- Started events after restart

Updates:
- Add test to test/extended/node/node_e2e/node.go
- Document test in test/extended/node/README.md

Relates: https://issues.redhat.com/browse/OCPBUGS-44493
Signed-off-by: Bhargavi Gudi <BhargaviGudi@users.noreply.github.com>

Author: BhargaviGudi

test/extended/node/README.md

@@ -19,6 +19,7 @@ This directory contains OpenShift end-to-end tests for node-related features.
 - **image_volume.go** - Tests mounting container images as volumes in pods, including subPath and error handling
 - **node_swap.go** - Tests default kubelet swap settings (failSwapOn and swapBehavior) and rejection of user overrides
 - **zstd_chunked.go** - Tests building and running images with zstd:chunked compression format
+- **node_e2e/node.go** - Probe-level terminationGracePeriodSeconds (OCP-44493) - Tests configurable termination grace period for liveness and startup probes [Lifecycle:informing]
 
 ## Directory Structure
 

test/extended/node/node_e2e/node.go

@@ -4,15 +4,19 @@ import (
 	"context"
 	"fmt"
 	"path/filepath"
+	"strconv"
 	"strings"
 	"time"
 
 	g "github.com/onsi/ginkgo/v2"
 	o "github.com/onsi/gomega"
+	ote "github.com/openshift-eng/openshift-tests-extension/pkg/ginkgo"
 
 	configv1 "github.com/openshift/api/config/v1"
 	"github.com/openshift/origin/test/extended/imagepolicy"
+	corev1 "k8s.io/api/core/v1"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/apimachinery/pkg/util/intstr"
 	utilrand "k8s.io/apimachinery/pkg/util/rand"
 	"k8s.io/apimachinery/pkg/util/wait"
 	e2e "k8s.io/kubernetes/test/e2e/framework"
@@ -164,6 +168,284 @@ var _ = g.Describe("[sig-node] [Jira:Node/Kubelet] Kubelet, CRI-O, CPU manager",
 		e2e.Logf("/dev/fuse mount output: %s", output)
 		o.Expect(output).To(o.ContainSubstring("fuse"), "dev fuse is not mounted inside pod")
 	})
+
+	//author: minmli@redhat.com
+	//migrated from openshift-tests-private
+	//automates: https://issues.redhat.com/browse/OCPBUGS-44493
+	g.It("[OTP] add configurable terminationGracePeriodSeconds to liveness and startup probes [OCP-44493]", ote.Informing(), func() {
+		ctx := context.Background()
+		var err error
+
+		oc.SetupProject()
+		namespace := oc.Namespace()
+
+		// Helper function to parse duration string like "1m30s" or "45s" to seconds
+		parseDurationToSeconds := func(durationStr string) (int, error) {
+			var totalSeconds int
+			if strings.Contains(durationStr, "m") {
+				parts := strings.Split(durationStr, "m")
+				minutes, err := strconv.Atoi(parts[0])
+				if err != nil {
+					return 0, err
+				}
+				totalSeconds = minutes * 60
+				if len(parts) > 1 && strings.Contains(parts[1], "s") {
+					secStr := strings.TrimSuffix(parts[1], "s")
+					if secStr != "" {
+						seconds, err := strconv.Atoi(secStr)
+						if err != nil {
+							return 0, err
+						}
+						totalSeconds += seconds
+					}
+				}
+			} else if strings.Contains(durationStr, "s") {
+				secStr := strings.TrimSuffix(durationStr, "s")
+				seconds, err := strconv.Atoi(secStr)
+				if err != nil {
+					return 0, err
+				}
+				totalSeconds = seconds
+			}
+			return totalSeconds, nil
+		}
+
+		// Helper to verify probe termination period
+		verifyProbeTermination := func(podName, containerName string, expectedTerminationSec int) error {
+			return wait.PollUntilContextTimeout(ctx, 10*time.Second, 5*time.Minute, true, func(ctx context.Context) (bool, error) {
+				podDesc, err := oc.AsAdmin().WithoutNamespace().Run("describe").Args("pod", podName, "-n", namespace).Output()
+				if err != nil {
+					e2e.Logf("Error describing pod: %v", err)
+					return false, nil
+				}
+
+				// Look for probe failure (killing) and container restart events
+				// Event format: "Normal  Killing  <time>  kubelet  Container <name> failed <probe> probe, will be restarted"
+				// Event format: "Normal  Started  <time>  kubelet  Container started"
+				killingLine := ""
+				restartLine := ""
+
+				inEvents := false
+				for _, line := range strings.Split(podDesc, "\n") {
+					if strings.Contains(line, "Events:") {
+						inEvents = true
+						continue
+					}
+					if !inEvents {
+						continue
+					}
+
+					// Look for killing event with container name
+					if strings.Contains(line, "Killing") && strings.Contains(line, containerName) &&
+						strings.Contains(line, "failed") && strings.Contains(line, "probe") &&
+						strings.Contains(line, "will be restarted") {
+						killingLine = line
+					}
+					// Look for Started event after Killing
+					if killingLine != "" && strings.Contains(line, "Started") && strings.Contains(line, "Container started") {
+						restartLine = line
+						break
+					}
+				}
+
+				if killingLine == "" || restartLine == "" {
+					e2e.Logf("Waiting for probe failure (killing) and container restart events")
+					return false, nil
+				}
+
+				e2e.Logf("Killing event: %s", killingLine)
+				e2e.Logf("Restart event: %s", restartLine)
+
+				// Extract timestamps (format: "1m30s" or "45s")
+				// Event format: "Normal  Killing  2m30s  kubelet  Container..."
+				killingFields := strings.Fields(killingLine)
+				restartFields := strings.Fields(restartLine)
+				if len(killingFields) < 3 || len(restartFields) < 3 {
+					e2e.Logf("Unable to parse event timestamps")
+					return false, nil
+				}
+
+				killingTime := killingFields[2]
+				restartTime := restartFields[2]
+
+				killingSec, err := parseDurationToSeconds(killingTime)
+				if err != nil {
+					e2e.Logf("Error parsing killing time: %v", err)
+					return false, nil
+				}
+
+				restartSec, err := parseDurationToSeconds(restartTime)
+				if err != nil {
+					e2e.Logf("Error parsing restart time: %v", err)
+					return false, nil
+				}
+
+				// Time difference: killing happened earlier, restart happened later
+				// So we need to calculate how long between them
+				timeDiff := killingSec - restartSec
+				e2e.Logf("Time difference: %d seconds (expected: %d ±10 seconds)", timeDiff, expectedTerminationSec)
+
+				// Allow range: [expectedTerminationSec-3, expectedTerminationSec+10]
+				if timeDiff >= (expectedTerminationSec-3) && timeDiff <= (expectedTerminationSec+10) {
+					e2e.Logf("Termination grace period check passed")
+					return true, nil
+				}
+
+				e2e.Logf("Time difference %d is outside expected range [%d, %d]", timeDiff, expectedTerminationSec-3, expectedTerminationSec+10)
+				return false, nil
+			})
+		}
+
+		g.By("Test liveness probe with probe-level terminationGracePeriodSeconds")
+		livenessPod := &corev1.Pod{
+			ObjectMeta: metav1.ObjectMeta{
+				Name:      "liveness-probe",
+				Namespace: namespace,
+			},
+			Spec: corev1.PodSpec{
+				TerminationGracePeriodSeconds: &[]int64{60}[0],
+				SecurityContext: &corev1.PodSecurityContext{
+					RunAsNonRoot: &[]bool{true}[0],
+					SeccompProfile: &corev1.SeccompProfile{
+						Type: corev1.SeccompProfileTypeRuntimeDefault,
+					},
+				},
+				Containers: []corev1.Container{
+					{
+						Name:  "test",
+						Image: "quay.io/openshifttest/nginx-alpine@sha256:04f316442d48ba60e3ea0b5a67eb89b0b667abf1c198a3d0056ca748736336a0",
+						SecurityContext: &corev1.SecurityContext{
+							AllowPrivilegeEscalation: &[]bool{false}[0],
+							Capabilities: &corev1.Capabilities{
+								Drop: []corev1.Capability{"ALL"},
+							},
+						},
+						Command: []string{"sh", "-c", "sleep 100000000"},
+						Ports: []corev1.ContainerPort{
+							{ContainerPort: 8080},
+						},
+						LivenessProbe: &corev1.Probe{
+							ProbeHandler: corev1.ProbeHandler{
+								HTTPGet: &corev1.HTTPGetAction{
+									Path: "/healthz",
+									Port: intstr.FromInt(8080),
+								},
+							},
+							FailureThreshold:              1,
+							PeriodSeconds:                 60,
+							TerminationGracePeriodSeconds: &[]int64{10}[0],
+						},
+					},
+				},
+			},
+		}
+
+		_, err = oc.KubeClient().CoreV1().Pods(namespace).Create(ctx, livenessPod, metav1.CreateOptions{})
+		o.Expect(err).NotTo(o.HaveOccurred(), "failed to create liveness probe pod")
+
+		err = verifyProbeTermination("liveness-probe", "test", 10)
+		o.Expect(err).NotTo(o.HaveOccurred(), "liveness probe termination grace period not honored")
+
+		g.By("Test startup probe with probe-level terminationGracePeriodSeconds")
+		startupPod := &corev1.Pod{
+			ObjectMeta: metav1.ObjectMeta{
+				Name:      "startup-probe",
+				Namespace: namespace,
+			},
+			Spec: corev1.PodSpec{
+				TerminationGracePeriodSeconds: &[]int64{60}[0],
+				SecurityContext: &corev1.PodSecurityContext{
+					RunAsNonRoot: &[]bool{true}[0],
+					SeccompProfile: &corev1.SeccompProfile{
+						Type: corev1.SeccompProfileTypeRuntimeDefault,
+					},
+				},
+				Containers: []corev1.Container{
+					{
+						Name:  "teststartup",
+						Image: "quay.io/openshifttest/nginx-alpine@sha256:04f316442d48ba60e3ea0b5a67eb89b0b667abf1c198a3d0056ca748736336a0",
+						SecurityContext: &corev1.SecurityContext{
+							AllowPrivilegeEscalation: &[]bool{false}[0],
+							Capabilities: &corev1.Capabilities{
+								Drop: []corev1.Capability{"ALL"},
+							},
+						},
+						Command: []string{"sh", "-c", "sleep 100000000"},
+						Ports: []corev1.ContainerPort{
+							{ContainerPort: 8080},
+						},
+						StartupProbe: &corev1.Probe{
+							ProbeHandler: corev1.ProbeHandler{
+								HTTPGet: &corev1.HTTPGetAction{
+									Path: "/healthz",
+									Port: intstr.FromInt(8080),
+								},
+							},
+							FailureThreshold:              1,
+							PeriodSeconds:                 60,
+							TerminationGracePeriodSeconds: &[]int64{10}[0],
+						},
+					},
+				},
+			},
+		}
+
+		_, err = oc.KubeClient().CoreV1().Pods(namespace).Create(ctx, startupPod, metav1.CreateOptions{})
+		o.Expect(err).NotTo(o.HaveOccurred(), "failed to create startup probe pod")
+
+		err = verifyProbeTermination("startup-probe", "teststartup", 10)
+		o.Expect(err).NotTo(o.HaveOccurred(), "startup probe termination grace period not honored")
+
+		g.By("Test liveness probe without probe-level terminationGracePeriodSeconds (should use pod-level)")
+		livenessPodNoProbeTerm := &corev1.Pod{
+			ObjectMeta: metav1.ObjectMeta{
+				Name:      "liveness-probe-no-term",
+				Namespace: namespace,
+			},
+			Spec: corev1.PodSpec{
+				TerminationGracePeriodSeconds: &[]int64{60}[0],
+				SecurityContext: &corev1.PodSecurityContext{
+					RunAsNonRoot: &[]bool{true}[0],
+					SeccompProfile: &corev1.SeccompProfile{
+						Type: corev1.SeccompProfileTypeRuntimeDefault,
+					},
+				},
+				Containers: []corev1.Container{
+					{
+						Name:  "test",
+						Image: "quay.io/openshifttest/nginx-alpine@sha256:04f316442d48ba60e3ea0b5a67eb89b0b667abf1c198a3d0056ca748736336a0",
+						SecurityContext: &corev1.SecurityContext{
+							AllowPrivilegeEscalation: &[]bool{false}[0],
+							Capabilities: &corev1.Capabilities{
+								Drop: []corev1.Capability{"ALL"},
+							},
+						},
+						Command: []string{"sh", "-c", "sleep 100000000"},
+						Ports: []corev1.ContainerPort{
+							{ContainerPort: 8080},
+						},
+						LivenessProbe: &corev1.Probe{
+							ProbeHandler: corev1.ProbeHandler{
+								HTTPGet: &corev1.HTTPGetAction{
+									Path: "/healthz",
+									Port: intstr.FromInt(8080),
+								},
+							},
+							FailureThreshold: 1,
+							PeriodSeconds:    60,
+							// No TerminationGracePeriodSeconds - should use pod-level (60s)
+						},
+					},
+				},
+			},
+		}
+
+		_, err = oc.KubeClient().CoreV1().Pods(namespace).Create(ctx, livenessPodNoProbeTerm, metav1.CreateOptions{})
+		o.Expect(err).NotTo(o.HaveOccurred(), "failed to create liveness probe pod without probe termination")
+
+		err = verifyProbeTermination("liveness-probe-no-term", "test", 60)
+		o.Expect(err).NotTo(o.HaveOccurred(), "liveness probe should use pod-level termination grace period when probe-level not set")
+	})
 })
 
 // author: asahay@redhat.com