feat: Add Global Default Termination Grace Period Variable (#2088)

Author: engedaam

pkg/controllers/provisioning/scheduling/nodeclaimtemplate.go

@@ -30,6 +30,11 @@ import (
 	"sigs.k8s.io/karpenter/pkg/scheduling"
 )
 
+// DefaultTerminationGracePeriod is used as runtime defaulting for TerminationGracePeriod on the NodeClaim
+// This would be a mechanism to allow cloud providers to enforce a TerminationGracePeriod on all node
+// provisioned by Karpenter
+var DefaultTerminationGracePeriod *metav1.Duration = nil
+
 // MaxInstanceTypes is a constant that restricts the number of instance types to be sent for launch. Note that this
 // is intentionally changed to var just to help in testing the code.
 var MaxInstanceTypes = 60
@@ -93,5 +98,9 @@ func (i *NodeClaimTemplate) ToNodeClaim() *v1.NodeClaim {
 		Spec: i.Spec,
 	}
 	nc.Spec.Requirements = i.Requirements.NodeSelectorRequirements()
+	if nc.Spec.TerminationGracePeriod == nil {
+		nc.Spec.TerminationGracePeriod = DefaultTerminationGracePeriod
+	}
+
 	return nc
 }

pkg/controllers/provisioning/suite_test.go

@@ -98,6 +98,7 @@ var _ = BeforeEach(func() {
 	}
 	fakeClock.SetTime(time.Now())
 	state.PodSchedulingDecisionSeconds.Reset()
+	pscheduling.DefaultTerminationGracePeriod = nil
 })
 
 var _ = AfterSuite(func() {
@@ -220,6 +221,42 @@ var _ = Describe("Provisioning", func() {
 		Expect(len(nodes.Items)).To(Equal(1))
 		ExpectScheduled(ctx, env.Client, pod)
 	})
+	It("should expect nodeclaim terminationGracePeriod to be the global value when the nodepool terminationGracePeriod is not set", func() {
+		nodePool := test.NodePool()
+		pscheduling.DefaultTerminationGracePeriod = &metav1.Duration{Duration: 98 * time.Hour}
+		ExpectApplied(ctx, env.Client, nodePool)
+		pod := test.UnschedulablePod()
+		ExpectProvisioned(ctx, env.Client, cluster, cloudProvider, prov, pod)
+		ExpectScheduled(ctx, env.Client, pod)
+		nodeClaims := &v1.NodeClaimList{}
+		Expect(env.Client.List(ctx, nodeClaims)).To(Succeed())
+		Expect(len(nodeClaims.Items)).To(Equal(1))
+		Expect(nodeClaims.Items[0].Spec.TerminationGracePeriod.Duration).To(BeNumerically("==", 98*time.Hour))
+	})
+	It("should expect nodeclaim terminationGracePeriod to be the nil when the nodepool terminationGracePeriod and global terminationGracePeriod is not set", func() {
+		nodePool := test.NodePool()
+		ExpectApplied(ctx, env.Client, nodePool)
+		pod := test.UnschedulablePod()
+		ExpectProvisioned(ctx, env.Client, cluster, cloudProvider, prov, pod)
+		ExpectScheduled(ctx, env.Client, pod)
+		nodeClaims := &v1.NodeClaimList{}
+		Expect(env.Client.List(ctx, nodeClaims)).To(Succeed())
+		Expect(len(nodeClaims.Items)).To(Equal(1))
+		Expect(nodeClaims.Items[0].Spec.TerminationGracePeriod).To(BeNil())
+	})
+	It("should respect terminationGracePeriod value set on the nodePool over global terminationGracePeriod", func() {
+		nodePool := test.NodePool()
+		nodePool.Spec.Template.Spec.TerminationGracePeriod = &metav1.Duration{Duration: 223 * time.Hour}
+		pscheduling.DefaultTerminationGracePeriod = &metav1.Duration{Duration: 47 * time.Hour}
+		ExpectApplied(ctx, env.Client, nodePool)
+		pod := test.UnschedulablePod()
+		ExpectProvisioned(ctx, env.Client, cluster, cloudProvider, prov, pod)
+		ExpectScheduled(ctx, env.Client, pod)
+		nodeClaims := &v1.NodeClaimList{}
+		Expect(env.Client.List(ctx, nodeClaims)).To(Succeed())
+		Expect(len(nodeClaims.Items)).To(Equal(1))
+		Expect(nodeClaims.Items[0].Spec.TerminationGracePeriod.Duration).To(BeNumerically("==", 223*time.Hour))
+	})
 	It("should ignore NodePools that are deleting", func() {
 		nodePool := test.NodePool()
 		ExpectApplied(ctx, env.Client, nodePool)