|
| 1 | +package tests |
| 2 | + |
| 3 | +import ( |
| 4 | + "context" |
| 5 | + "strings" |
| 6 | + "time" |
| 7 | + |
| 8 | + metav1 "k8s.io/apimachinery/pkg/apis/meta/v1" |
| 9 | + |
| 10 | + netattdefv1 "github.com/k8snetworkplumbingwg/network-attachment-definition-client/pkg/apis/k8s.cni.cncf.io/v1" |
| 11 | + |
| 12 | + sriovv1 "github.com/k8snetworkplumbingwg/sriov-network-operator/api/v1" |
| 13 | + "github.com/k8snetworkplumbingwg/sriov-network-operator/pkg/consts" |
| 14 | + "github.com/k8snetworkplumbingwg/sriov-network-operator/test/util/cluster" |
| 15 | + "github.com/k8snetworkplumbingwg/sriov-network-operator/test/util/discovery" |
| 16 | + "github.com/k8snetworkplumbingwg/sriov-network-operator/test/util/namespaces" |
| 17 | + "github.com/k8snetworkplumbingwg/sriov-network-operator/test/util/nodes" |
| 18 | + |
| 19 | + . "github.com/onsi/ginkgo/v2" |
| 20 | + . "github.com/onsi/gomega" |
| 21 | + corev1 "k8s.io/api/core/v1" |
| 22 | + k8serrors "k8s.io/apimachinery/pkg/api/errors" |
| 23 | + runtimeclient "sigs.k8s.io/controller-runtime/pkg/client" |
| 24 | +) |
| 25 | + |
| 26 | +var _ = Describe("[sriov] operator", Ordered, ContinueOnFailure, func() { |
| 27 | + Describe("No SriovNetworkNodePolicy", func() { |
| 28 | + Context("SR-IOV network config daemon can be set by nodeselector", func() { |
| 29 | + // 26186 |
| 30 | + It("Should schedule the config daemon on selected nodes", func() { |
| 31 | + if discovery.Enabled() { |
| 32 | + Skip("Test unsuitable to be run in discovery mode") |
| 33 | + } |
| 34 | + |
| 35 | + By("Checking that a daemon is scheduled on each worker node") |
| 36 | + Eventually(func() bool { |
| 37 | + return daemonsScheduledOnNodes("node-role.kubernetes.io/worker=") |
| 38 | + }, 3*time.Minute, 1*time.Second).Should(Equal(true)) |
| 39 | + |
| 40 | + By("Labeling one worker node with the label needed for the daemon") |
| 41 | + allNodes, err := clients.CoreV1Interface.Nodes().List(context.Background(), metav1.ListOptions{ |
| 42 | + LabelSelector: "node-role.kubernetes.io/worker", |
| 43 | + }) |
| 44 | + Expect(err).ToNot(HaveOccurred()) |
| 45 | + |
| 46 | + selectedNodes, err := nodes.MatchingOptionalSelector(clients, allNodes.Items) |
| 47 | + Expect(err).ToNot(HaveOccurred()) |
| 48 | + |
| 49 | + Expect(len(selectedNodes)).To(BeNumerically(">", 0), "There must be at least one worker") |
| 50 | + candidate := selectedNodes[0] |
| 51 | + candidate.Labels["sriovenabled"] = "true" |
| 52 | + _, err = clients.CoreV1Interface.Nodes().Update(context.Background(), &candidate, metav1.UpdateOptions{}) |
| 53 | + Expect(err).ToNot(HaveOccurred()) |
| 54 | + |
| 55 | + By("Setting the node selector for each daemon") |
| 56 | + cfg := sriovv1.SriovOperatorConfig{} |
| 57 | + err = clients.Get(context.TODO(), runtimeclient.ObjectKey{ |
| 58 | + Name: "default", |
| 59 | + Namespace: operatorNamespace, |
| 60 | + }, &cfg) |
| 61 | + Expect(err).ToNot(HaveOccurred()) |
| 62 | + cfg.Spec.ConfigDaemonNodeSelector = map[string]string{ |
| 63 | + "sriovenabled": "true", |
| 64 | + } |
| 65 | + Eventually(func() error { |
| 66 | + return clients.Update(context.TODO(), &cfg) |
| 67 | + }, 1*time.Minute, 5*time.Second).ShouldNot(HaveOccurred()) |
| 68 | + |
| 69 | + By("Checking that a daemon is scheduled only on selected node") |
| 70 | + Eventually(func() bool { |
| 71 | + return !daemonsScheduledOnNodes("sriovenabled!=true") && |
| 72 | + daemonsScheduledOnNodes("sriovenabled=true") |
| 73 | + }, 1*time.Minute, 1*time.Second).Should(Equal(true)) |
| 74 | + |
| 75 | + By("Restoring the node selector for daemons") |
| 76 | + err = clients.Get(context.TODO(), runtimeclient.ObjectKey{ |
| 77 | + Name: "default", |
| 78 | + Namespace: operatorNamespace, |
| 79 | + }, &cfg) |
| 80 | + Expect(err).ToNot(HaveOccurred()) |
| 81 | + cfg.Spec.ConfigDaemonNodeSelector = map[string]string{} |
| 82 | + Eventually(func() error { |
| 83 | + return clients.Update(context.TODO(), &cfg) |
| 84 | + }, 1*time.Minute, 5*time.Second).ShouldNot(HaveOccurred()) |
| 85 | + |
| 86 | + By("Checking that a daemon is scheduled on each worker node") |
| 87 | + Eventually(func() bool { |
| 88 | + return daemonsScheduledOnNodes("node-role.kubernetes.io/worker") |
| 89 | + }, 1*time.Minute, 1*time.Second).Should(Equal(true)) |
| 90 | + }) |
| 91 | + }) |
| 92 | + |
| 93 | + Context("LogLevel affects operator's logs", func() { |
| 94 | + It("when set to 0 no lifecycle logs are present", func() { |
| 95 | + if discovery.Enabled() { |
| 96 | + Skip("Test unsuitable to be run in discovery mode") |
| 97 | + } |
| 98 | + |
| 99 | + initialLogLevelValue := getOperatorConfigLogLevel() |
| 100 | + DeferCleanup(func() { |
| 101 | + By("Restore LogLevel to its initial value") |
| 102 | + setOperatorConfigLogLevel(initialLogLevelValue) |
| 103 | + }) |
| 104 | + |
| 105 | + initialDisableDrain, err := cluster.GetNodeDrainState(clients, operatorNamespace) |
| 106 | + Expect(err).ToNot(HaveOccurred()) |
| 107 | + |
| 108 | + DeferCleanup(func() { |
| 109 | + By("Restore DisableDrain to its initial value") |
| 110 | + Eventually(func() error { |
| 111 | + return cluster.SetDisableNodeDrainState(clients, operatorNamespace, initialDisableDrain) |
| 112 | + }, 1*time.Minute, 5*time.Second).ShouldNot(HaveOccurred()) |
| 113 | + }) |
| 114 | + |
| 115 | + By("Set operator LogLevel to 2") |
| 116 | + setOperatorConfigLogLevel(2) |
| 117 | + |
| 118 | + By("Flip DisableDrain to trigger operator activity") |
| 119 | + since := time.Now().Add(-10 * time.Second) |
| 120 | + Eventually(func() error { |
| 121 | + return cluster.SetDisableNodeDrainState(clients, operatorNamespace, !initialDisableDrain) |
| 122 | + }, 1*time.Minute, 5*time.Second).ShouldNot(HaveOccurred()) |
| 123 | + |
| 124 | + By("Assert logs contains verbose output") |
| 125 | + Eventually(func(g Gomega) { |
| 126 | + logs := getOperatorLogs(since) |
| 127 | + g.Expect(logs).To( |
| 128 | + ContainElement(And( |
| 129 | + ContainSubstring("Reconciling SriovOperatorConfig"), |
| 130 | + )), |
| 131 | + ) |
| 132 | + |
| 133 | + // Should contain verbose logging |
| 134 | + g.Expect(logs).To( |
| 135 | + ContainElement( |
| 136 | + ContainSubstring("Start to sync webhook objects"), |
| 137 | + ), |
| 138 | + ) |
| 139 | + }, 1*time.Minute, 5*time.Second).Should(Succeed()) |
| 140 | + |
| 141 | + By("Reduce operator LogLevel to 0") |
| 142 | + setOperatorConfigLogLevel(0) |
| 143 | + |
| 144 | + By("Flip DisableDrain again to trigger operator activity") |
| 145 | + since = time.Now().Add(-10 * time.Second) |
| 146 | + Eventually(func() error { |
| 147 | + return cluster.SetDisableNodeDrainState(clients, operatorNamespace, initialDisableDrain) |
| 148 | + }, 1*time.Minute, 5*time.Second).ShouldNot(HaveOccurred()) |
| 149 | + |
| 150 | + By("Assert logs contains less operator activity") |
| 151 | + Eventually(func(g Gomega) { |
| 152 | + logs := getOperatorLogs(since) |
| 153 | + |
| 154 | + // time only contains sec, but we can have race here that in the same sec there was a sync |
| 155 | + afterLogs := []string{} |
| 156 | + found := false |
| 157 | + for _, log := range logs { |
| 158 | + if found { |
| 159 | + afterLogs = append(afterLogs, log) |
| 160 | + } |
| 161 | + if strings.Contains(log, "{\"new-level\": 0, \"current-level\": 2}") { |
| 162 | + found = true |
| 163 | + } |
| 164 | + } |
| 165 | + g.Expect(found).To(BeTrue()) |
| 166 | + g.Expect(afterLogs).To( |
| 167 | + ContainElement(And( |
| 168 | + ContainSubstring("Reconciling SriovOperatorConfig"), |
| 169 | + )), |
| 170 | + ) |
| 171 | + |
| 172 | + // Should not contain verbose logging |
| 173 | + g.Expect(afterLogs).ToNot( |
| 174 | + ContainElement( |
| 175 | + ContainSubstring("Start to sync webhook objects"), |
| 176 | + ), |
| 177 | + ) |
| 178 | + }, 3*time.Minute, 5*time.Second).Should(Succeed()) |
| 179 | + }) |
| 180 | + }) |
| 181 | + |
| 182 | + DescribeTable("should gracefully restart quickly", func(webookEnabled bool) { |
| 183 | + DeferCleanup(setSriovOperatorSpecFlag(operatorNetworkInjectorFlag, webookEnabled)) |
| 184 | + DeferCleanup(setSriovOperatorSpecFlag(operatorWebhookFlag, webookEnabled)) |
| 185 | + |
| 186 | + // This test case ensure leader election process runs smoothly when the operator's pod is restarted. |
| 187 | + oldLease, err := clients.CoordinationV1Interface.Leases(operatorNamespace).Get(context.Background(), consts.LeaderElectionID, metav1.GetOptions{}) |
| 188 | + if k8serrors.IsNotFound(err) { |
| 189 | + Skip("Leader Election is not enabled on the cluster. Skipping") |
| 190 | + } |
| 191 | + Expect(err).ToNot(HaveOccurred()) |
| 192 | + |
| 193 | + oldOperatorPod := getOperatorPod() |
| 194 | + |
| 195 | + By("Delete the operator's pod") |
| 196 | + deleteOperatorPod() |
| 197 | + |
| 198 | + By("Wait the new operator's pod to start") |
| 199 | + Eventually(func(g Gomega) { |
| 200 | + newOperatorPod := getOperatorPod() |
| 201 | + Expect(newOperatorPod.Name).ToNot(Equal(oldOperatorPod.Name)) |
| 202 | + Expect(newOperatorPod.Status.Phase).To(Equal(corev1.PodRunning)) |
| 203 | + }, 45*time.Second, 5*time.Second) |
| 204 | + |
| 205 | + By("Assert the new operator's pod acquire the lease before 30 seconds") |
| 206 | + Eventually(func(g Gomega) { |
| 207 | + newLease, err := clients.CoordinationV1Interface.Leases(operatorNamespace).Get(context.Background(), consts.LeaderElectionID, metav1.GetOptions{}) |
| 208 | + g.Expect(err).ToNot(HaveOccurred()) |
| 209 | + |
| 210 | + g.Expect(newLease.Spec.HolderIdentity).ToNot(Equal(oldLease.Spec.HolderIdentity)) |
| 211 | + }, 30*time.Second, 5*time.Second).Should(Succeed()) |
| 212 | + }, |
| 213 | + Entry("webhooks enabled", true), |
| 214 | + Entry("webhooks disabled", true), |
| 215 | + ) |
| 216 | + |
| 217 | + Context("SriovNetworkMetricsExporter", func() { |
| 218 | + BeforeEach(func() { |
| 219 | + if discovery.Enabled() { |
| 220 | + Skip("Test unsuitable to be run in discovery mode") |
| 221 | + } |
| 222 | + |
| 223 | + initialValue := isFeatureFlagEnabled("metricsExporter") |
| 224 | + DeferCleanup(func() { |
| 225 | + By("Restoring initial feature flag value") |
| 226 | + setFeatureFlag("metricsExporter", initialValue) |
| 227 | + }) |
| 228 | + |
| 229 | + By("Enabling `metricsExporter` feature flag") |
| 230 | + setFeatureFlag("metricsExporter", true) |
| 231 | + }) |
| 232 | + |
| 233 | + It("should be deployed if the feature gate is enabled", func() { |
| 234 | + By("Checking that a daemon is scheduled on selected node") |
| 235 | + Eventually(func() bool { |
| 236 | + return isDaemonsetScheduledOnNodes("node-role.kubernetes.io/worker", "app=sriov-network-metrics-exporter") |
| 237 | + }).WithTimeout(time.Minute).WithPolling(time.Second).Should(Equal(true)) |
| 238 | + }) |
| 239 | + |
| 240 | + It("should deploy ServiceMonitor if the Prometheus operator is installed", func() { |
| 241 | + _, err := clients.ServiceMonitors(operatorNamespace).List(context.Background(), metav1.ListOptions{}) |
| 242 | + if k8serrors.IsNotFound(err) { |
| 243 | + Skip("Prometheus operator not available in the cluster") |
| 244 | + } |
| 245 | + |
| 246 | + By("Checking ServiceMonitor is deployed if needed") |
| 247 | + Eventually(func(g Gomega) { |
| 248 | + _, err := clients.ServiceMonitors(operatorNamespace).Get(context.Background(), "sriov-network-metrics-exporter", metav1.GetOptions{}) |
| 249 | + g.Expect(err).ToNot(HaveOccurred()) |
| 250 | + }).WithTimeout(time.Minute).WithPolling(time.Second).Should(Succeed()) |
| 251 | + }) |
| 252 | + |
| 253 | + It("should remove ServiceMonitor when the feature is turned off", func() { |
| 254 | + setFeatureFlag("metricsExporter", false) |
| 255 | + Eventually(func(g Gomega) { |
| 256 | + _, err := clients.ServiceMonitors(operatorNamespace).Get(context.Background(), "sriov-network-metrics-exporter", metav1.GetOptions{}) |
| 257 | + g.Expect(k8serrors.IsNotFound(err)).To(BeTrue()) |
| 258 | + }).WithTimeout(time.Minute).WithPolling(time.Second).Should(Succeed()) |
| 259 | + }) |
| 260 | + }) |
| 261 | + |
| 262 | + Context("Namespaced network objects", func() { |
| 263 | + BeforeAll(func() { |
| 264 | + |
| 265 | + }) |
| 266 | + |
| 267 | + DescribeTable("can be create in every namespaces", func(object runtimeclient.Object) { |
| 268 | + err := clients.Create(context.Background(), object) |
| 269 | + Expect(err).ToNot(HaveOccurred()) |
| 270 | + |
| 271 | + waitForNetAttachDef(object.GetName(), object.GetNamespace()) |
| 272 | + |
| 273 | + err = clients.Delete(context.Background(), object) |
| 274 | + Expect(err).ToNot(HaveOccurred()) |
| 275 | + |
| 276 | + Eventually(func() bool { |
| 277 | + netAttDef := &netattdefv1.NetworkAttachmentDefinition{} |
| 278 | + err := clients.Get(context.Background(), runtimeclient.ObjectKey{Name: object.GetName(), Namespace: object.GetNamespace()}, netAttDef) |
| 279 | + return err != nil && k8serrors.IsNotFound(err) |
| 280 | + }, 2*time.Minute, 10*time.Second).Should(BeTrue()) |
| 281 | + }, |
| 282 | + Entry("SriovNetwork", &sriovv1.SriovNetwork{ObjectMeta: metav1.ObjectMeta{Name: "sriovnet1", Namespace: namespaces.Test}}), |
| 283 | + Entry("SriovIBNetwork", &sriovv1.SriovIBNetwork{ObjectMeta: metav1.ObjectMeta{Name: "sriovibnet1", Namespace: namespaces.Test}}), |
| 284 | + Entry("OVSNetwork", &sriovv1.OVSNetwork{ObjectMeta: metav1.ObjectMeta{Name: "ovsnet1", Namespace: namespaces.Test}}), |
| 285 | + ) |
| 286 | + |
| 287 | + DescribeTable("can NOT be in application namespace and have .Spec.NetworkNamespace != ''", func(object runtimeclient.Object) { |
| 288 | + err := clients.Create(context.Background(), object) |
| 289 | + Expect(err).To(HaveOccurred()) |
| 290 | + Expect(string(k8serrors.ReasonForError(err))). |
| 291 | + To(ContainSubstring(".Spec.NetworkNamespace field can't be specified if the resource is not in the ")) |
| 292 | + }, |
| 293 | + Entry("SriovNetwork", &sriovv1.SriovNetwork{ObjectMeta: metav1.ObjectMeta{Name: "sriovnet1", Namespace: namespaces.Test}, Spec: sriovv1.SriovNetworkSpec{NetworkNamespace: "default"}}), |
| 294 | + Entry("SriovIBNetwork", &sriovv1.SriovIBNetwork{ObjectMeta: metav1.ObjectMeta{Name: "sriovibnet1", Namespace: namespaces.Test}, Spec: sriovv1.SriovIBNetworkSpec{NetworkNamespace: "default"}}), |
| 295 | + Entry("OVSNetwork", &sriovv1.OVSNetwork{ObjectMeta: metav1.ObjectMeta{Name: "ovsnet1", Namespace: namespaces.Test}, Spec: sriovv1.OVSNetworkSpec{NetworkNamespace: "default"}}), |
| 296 | + ) |
| 297 | + }) |
| 298 | + }) |
| 299 | +}) |
0 commit comments