From e1e38148bb2e81a84e4c35598037619d020fe323 Mon Sep 17 00:00:00 2001
From: Michal Wozniak <michalwozniak@google.com>
Date: Tue, 25 Apr 2023 14:02:39 +0200
Subject: [PATCH] Add DisruptionTarget condition when preempting for critical
 pod

---
 pkg/kubelet/preemption/preemption.go | 11 +++++++
 test/e2e_node/critical_pod_test.go   | 44 ++++++++++++++++++++++++++++
 2 files changed, 55 insertions(+)

diff --git a/pkg/kubelet/preemption/preemption.go b/pkg/kubelet/preemption/preemption.go
index 5f0fb5e03c0..e4d0cbd931b 100644
--- a/pkg/kubelet/preemption/preemption.go
+++ b/pkg/kubelet/preemption/preemption.go
@@ -21,10 +21,13 @@ import (
 	"math"
 
 	v1 "k8s.io/api/core/v1"
+	utilfeature "k8s.io/apiserver/pkg/util/feature"
 	"k8s.io/client-go/tools/record"
 	"k8s.io/klog/v2"
+	podutil "k8s.io/kubernetes/pkg/api/v1/pod"
 	"k8s.io/kubernetes/pkg/api/v1/resource"
 	v1qos "k8s.io/kubernetes/pkg/apis/core/v1/helper/qos"
+	"k8s.io/kubernetes/pkg/features"
 	"k8s.io/kubernetes/pkg/kubelet/events"
 	"k8s.io/kubernetes/pkg/kubelet/eviction"
 	"k8s.io/kubernetes/pkg/kubelet/lifecycle"
@@ -103,6 +106,14 @@ func (c *CriticalPodAdmissionHandler) evictPodsToFreeRequests(admitPod *v1.Pod,
 			status.Phase = v1.PodFailed
 			status.Reason = events.PreemptContainer
 			status.Message = message
+			if utilfeature.DefaultFeatureGate.Enabled(features.PodDisruptionConditions) {
+				podutil.UpdatePodCondition(status, &v1.PodCondition{
+					Type:    v1.DisruptionTarget,
+					Status:  v1.ConditionTrue,
+					Reason:  v1.PodReasonTerminationByKubelet,
+					Message: "Pod was preempted by Kubelet to accommodate a critical pod.",
+				})
+			}
 		})
 		if err != nil {
 			klog.ErrorS(err, "Failed to evict pod", "pod", klog.KObj(pod))
diff --git a/test/e2e_node/critical_pod_test.go b/test/e2e_node/critical_pod_test.go
index fa700fdb084..088640117ac 100644
--- a/test/e2e_node/critical_pod_test.go
+++ b/test/e2e_node/critical_pod_test.go
@@ -23,6 +23,7 @@ import (
 	v1 "k8s.io/api/core/v1"
 	"k8s.io/apimachinery/pkg/api/resource"
 	metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
+	"k8s.io/klog/v2"
 	kubeapi "k8s.io/kubernetes/pkg/apis/core"
 	"k8s.io/kubernetes/pkg/apis/scheduling"
 	kubelettypes "k8s.io/kubernetes/pkg/kubelet/types"
@@ -87,6 +88,49 @@ var _ = SIGDescribe("CriticalPod [Serial] [Disruptive] [NodeFeature:CriticalPod]
 				}
 			}
 		})
+
+		ginkgo.It("should add DisruptionTarget condition to the preempted pod [NodeFeature:PodDisruptionConditions]", func(ctx context.Context) {
+			// because adminssion Priority enable, If the priority class is not found, the Pod is rejected.
+			node := getNodeName(ctx, f)
+			nonCriticalGuaranteed := getTestPod(false, guaranteedPodName, v1.ResourceRequirements{
+				Requests: v1.ResourceList{
+					v1.ResourceCPU:    resource.MustParse("100m"),
+					v1.ResourceMemory: resource.MustParse("100Mi"),
+				},
+				Limits: v1.ResourceList{
+					v1.ResourceCPU:    resource.MustParse("100m"),
+					v1.ResourceMemory: resource.MustParse("100Mi"),
+				},
+			}, node)
+
+			criticalPod := getTestPod(true, criticalPodName, v1.ResourceRequirements{
+				// request the entire resource capacity of the node, so that
+				// admitting this pod requires the other pod to be preempted
+				Requests: getNodeCPUAndMemoryCapacity(ctx, f),
+			}, node)
+			criticalPod.Namespace = kubeapi.NamespaceSystem
+
+			ginkgo.By(fmt.Sprintf("create the non-critical pod %q", klog.KObj(nonCriticalGuaranteed)))
+			e2epod.NewPodClient(f).CreateSync(ctx, nonCriticalGuaranteed)
+
+			ginkgo.By(fmt.Sprintf("create the critical pod %q", klog.KObj(criticalPod)))
+			e2epod.PodClientNS(f, kubeapi.NamespaceSystem).Create(ctx, criticalPod)
+
+			ginkgo.By(fmt.Sprintf("await for the critical pod %q to be ready", klog.KObj(criticalPod)))
+			err := e2epod.WaitForPodNameRunningInNamespace(ctx, f.ClientSet, criticalPod.Name, kubeapi.NamespaceSystem)
+			framework.ExpectNoError(err, "Failed to await for the pod to be running: %q", klog.KObj(criticalPod))
+
+			// Check that non-critical pods other than the besteffort have been evicted
+			updatedPodList, err := f.ClientSet.CoreV1().Pods(f.Namespace.Name).List(ctx, metav1.ListOptions{})
+			framework.ExpectNoError(err)
+			for _, p := range updatedPodList.Items {
+				ginkgo.By(fmt.Sprintf("verify that the non-critical pod %q is preempted and has the DisruptionTarget condition", klog.KObj(&p)))
+				framework.ExpectEqual(p.Status.Phase, v1.PodSucceeded, fmt.Sprintf("pod: %v should be preempted with status: %#v", p.Name, p.Status))
+				if condition := e2epod.FindPodConditionByType(&p.Status, v1.DisruptionTarget); condition == nil {
+					framework.Failf("pod %q should have the condition: %q, pod status: %v", klog.KObj(&p), v1.DisruptionTarget, p.Status)
+				}
+			}
+		})
 		ginkgo.AfterEach(func(ctx context.Context) {
 			// Delete Pods
 			e2epod.NewPodClient(f).DeleteSync(ctx, guaranteedPodName, metav1.DeleteOptions{}, e2epod.DefaultPodDeletionTimeout)