feat: add evictAfterOOMThreshold per vpa

Signed-off-by: Omer Aplatony <[email protected]>

Author: omerap12

vertical-pod-autoscaler/charts/vertical-pod-autoscaler/crds/vpa-v1-crd-gen.yaml

@@ -254,6 +254,14 @@ spec:
     - jsonPath: .metadata.creationTimestamp
       name: Age
       type: date
+    - jsonPath: .spec.updatePolicy.minReplicas
+      name: MinReplicas
+      priority: 1
+      type: integer
+    - jsonPath: .spec.updatePolicy.evictAfterOOMThreshold
+      name: OOMThreshold
+      priority: 1
+      type: string
     name: v1
     schema:
       openAPIV3Schema:
@@ -425,6 +433,14 @@ spec:
                   If not specified, all fields in the `PodUpdatePolicy` are set to their
                   default values.
                 properties:
+                  evictAfterOOMThreshold:
+                    description: |-
+                      EvictAfterOOMThreshold specifies the time to wait after an OOM event before
+                      considering the pod for eviction. Pods that have OOMed in less than this threshold
+                      since start will be evicted.
+                    format: duration
+                    pattern: ^([0-9]+(\.[0-9]+)?(ns|us|µs|ms|s|m|h))+$
+                    type: string
                   evictionRequirements:
                     description: |-
                       EvictionRequirements is a list of EvictionRequirements that need to
@@ -478,6 +494,10 @@ spec:
                     - Auto
                     type: string
                 type: object
+                x-kubernetes-validations:
+                - message: evictAfterOOMThreshold must be greater than 0
+                  rule: '!has(self.evictAfterOOMThreshold) || duration(self.evictAfterOOMThreshold)
+                    >= duration(''0s'')'
             required:
             - targetRef
             type: object

vertical-pod-autoscaler/deploy/vpa-v1-crd-gen.yaml

@@ -254,6 +254,14 @@ spec:
     - jsonPath: .metadata.creationTimestamp
       name: Age
       type: date
+    - jsonPath: .spec.updatePolicy.minReplicas
+      name: MinReplicas
+      priority: 1
+      type: integer
+    - jsonPath: .spec.updatePolicy.evictAfterOOMThreshold
+      name: OOMThreshold
+      priority: 1
+      type: string
     name: v1
     schema:
       openAPIV3Schema:
@@ -425,6 +433,14 @@ spec:
                   If not specified, all fields in the `PodUpdatePolicy` are set to their
                   default values.
                 properties:
+                  evictAfterOOMThreshold:
+                    description: |-
+                      EvictAfterOOMThreshold specifies the time to wait after an OOM event before
+                      considering the pod for eviction. Pods that have OOMed in less than this threshold
+                      since start will be evicted.
+                    format: duration
+                    pattern: ^([0-9]+(\.[0-9]+)?(ns|us|µs|ms|s|m|h))+$
+                    type: string
                   evictionRequirements:
                     description: |-
                       EvictionRequirements is a list of EvictionRequirements that need to
@@ -478,6 +494,10 @@ spec:
                     - Auto
                     type: string
                 type: object
+                x-kubernetes-validations:
+                - message: evictAfterOOMThreshold must be greater than 0
+                  rule: '!has(self.evictAfterOOMThreshold) || duration(self.evictAfterOOMThreshold)
+                    >= duration(''0s'')'
             required:
             - targetRef
             type: object

vertical-pod-autoscaler/docs/flags.md

@@ -140,7 +140,7 @@ This document is auto-generated from the flag definitions in the VPA updater cod
 | `add-dir-header` |  |  | If true, adds the file directory to the header of the log messages |
 | `address` | string |  ":8943" | The address to expose Prometheus metrics.  |
 | `alsologtostderr` |  |  | log to standard error as well as files (no effect when -logtostderr=true) |
-| `evict-after-oom-threshold` |  |  10m0s | duration                              Evict pod that has OOMed in less than evict-after-oom-threshold since start.  |
+| `evict-after-oom-threshold` |  |  10m0s | duration                              The default duration to evict pod that has OOMed in less than evict-after-oom-threshold since start.  |
 | `eviction-rate-burst` | int |  1 | Burst of pods that can be evicted.  |
 | `eviction-rate-limit` | float |  | Number of pods that can be evicted per seconds. A rate limit set to 0 or -1 will disable<br>the rate limiter. (default -1) |
 | `eviction-tolerance` | float |  0.5 | Fraction of replica count that can be evicted for update, if more than one pod can be evicted.  |

vertical-pod-autoscaler/e2e/utils/common.go

@@ -70,6 +70,9 @@ var RecommenderLabels = map[string]string{"app": "vpa-recommender"}
 // HamsterLabels are labels of hamster app
 var HamsterLabels = map[string]string{"app": "hamster"}
 
+// OOMLabels are labels for OOM test pods
+var OOMLabels = map[string]string{"app": "oom-test"}
+
 // SIGDescribe adds sig-autoscaling tag to test description.
 // Takes args that are passed to ginkgo.Describe.
 func SIGDescribe(scenario, name string, args ...interface{}) bool {
@@ -226,7 +229,7 @@ func NewNHamstersDeployment(f *framework.Framework, n int) *appsv1.Deployment {
 		DefaultHamsterReplicas,                     /*replicas*/
 		HamsterLabels,                              /*podLabels*/
 		GetHamsterContainerNameByIndex(0),          /*imageName*/
-		"registry.k8s.io/ubuntu-slim:0.14",         /*image*/
+		"ubuntu:25.10",                             /*image*/
 		appsv1.RollingUpdateDeploymentStrategyType, /*strategyType*/
 	)
 	d.ObjectMeta.Namespace = f.Namespace.Name

vertical-pod-autoscaler/e2e/v1/actuation.go

@@ -203,7 +203,7 @@ var _ = ActuationSuiteE2eDescribe("Actuation", func() {
 		utils.InstallVPA(f, vpaCRD)
 
 		ginkgo.By("Waiting for pods to be evicted")
-		err = WaitForPodsEvicted(f, podList)
+		err = WaitForPodsEvicted(f, podList, utils.PollTimeout)
 		gomega.Expect(err).NotTo(gomega.HaveOccurred())
 	})
 
@@ -276,7 +276,7 @@ var _ = ActuationSuiteE2eDescribe("Actuation", func() {
 		}, restriction.DeferredResizeUpdateTimeout, 10*time.Second).Should(gomega.Succeed())
 
 		ginkgo.By("Waiting for pods to be evicted")
-		err = WaitForPodsEvicted(f, podList)
+		err = WaitForPodsEvicted(f, podList, utils.PollTimeout)
 		gomega.Expect(err).NotTo(gomega.HaveOccurred())
 	})
 
@@ -838,7 +838,7 @@ func testEvictsReplicatedPods(f *framework.Framework, controller *autoscaling.Cr
 	utils.InstallVPA(f, vpaCRD)
 
 	ginkgo.By("Waiting for pods to be evicted")
-	err = WaitForPodsEvicted(f, podList)
+	err = WaitForPodsEvicted(f, podList, utils.PollTimeout)
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
 }
 
@@ -898,7 +898,7 @@ func testEvictsSingletonPodWhenConfigured(f *framework.Framework, controller *au
 	utils.InstallVPA(f, vpaCRD)
 
 	ginkgo.By("Waiting for pods to be evicted")
-	err = WaitForPodsEvicted(f, podList)
+	err = WaitForPodsEvicted(f, podList, utils.PollTimeout)
 	gomega.Expect(err).NotTo(gomega.HaveOccurred())
 }