e2e: rewrite test 47674

The previous version was fragile and maintained several times already.
This fact means that the test should be rewritten and once the known bugs
are fixed, new tests will be added to cover the e2e full scenario.

The test goal is to verify that adjustments to NROP CR are done as
expected and reflected in the RTE daemonsets. Previously, this was done
via creating a new MCP for the test and associating to it one of the
worker nodes that were part of the initial NROP node groups. This was
causing a reboot when running on a cluster that is:
a. < 4.18 or
b. >= 4.18 and has the custom RTE selinux policy enabled

In the new version of the test, we keep checking the same functionality
while completely avoiding reboot. That is acheived by creating the new MCP,
keeping it empty (no nodes are associated to it) and appending it as a
second node group's selector. Then checks on the daemonsets are
performed and the only cleanup would be deleting the test mcp.

Signed-off-by: Shereen Haj <[email protected]>

internal/api/features/_topics.json

@@ -20,6 +20,7 @@
     "ngpoolname",
     "nganns",
     "metrics",
-    "nrtcrdanns"
+    "nrtcrdanns",
+    "tmp_b52958"
   ]
 }

test/e2e/serial/tests/configuration.go

@@ -47,6 +47,7 @@ import (
 	nrtv1alpha2 "github.com/k8stopologyawareschedwg/noderesourcetopology-api/pkg/apis/topology/v1alpha2"
 
 	configv1 "github.com/openshift/api/config/v1"
+	operatorv1 "github.com/openshift/api/operator/v1"
 	perfprof "github.com/openshift/cluster-node-tuning-operator/pkg/apis/performanceprofile/v2"
 	machineconfigv1 "github.com/openshift/machine-config-operator/pkg/apis/machineconfiguration.openshift.io/v1"
 
@@ -60,6 +61,7 @@ import (
 	e2ereslist "github.com/openshift-kni/numaresources-operator/internal/resourcelist"
 	"github.com/openshift-kni/numaresources-operator/internal/wait"
 	"github.com/openshift-kni/numaresources-operator/pkg/kubeletconfig"
+	"github.com/openshift-kni/numaresources-operator/pkg/objectnames"
 	"github.com/openshift-kni/numaresources-operator/pkg/status"
 	"github.com/openshift-kni/numaresources-operator/pkg/validation"
 	rteconfig "github.com/openshift-kni/numaresources-operator/rte/pkg/config"
@@ -138,6 +140,105 @@ var _ = Describe("[serial][disruptive] numaresources configuration management",
 	Context("cluster has at least one suitable node", func() {
 		timeout := 5 * time.Minute
 
+		It("[test_id:47674][images][tier2] should be able to modify the configurable values under the NUMAResourcesOperator CR", Label("images", "tier2"), Label("feature:tmp_b52958"), func(ctx context.Context) {
+			var nroOperObj nropv1.NUMAResourcesOperator
+			nroKey := objects.NROObjectKey()
+			Expect(fxt.Client.Get(ctx, nroKey, &nroOperObj)).To(Succeed())
+			initialNroOperObj := nroOperObj.DeepCopy()
+
+			testMCP := objects.TestMCP()
+			By(fmt.Sprintf("creating new MCP: %q", testMCP.Name))
+			// we must have this label in order to match other machine configs that are necessary for proper functionality
+			testMCP.Labels = map[string]string{"machineconfiguration.openshift.io/role": roleMCPTest}
+			testMCP.Spec.MachineConfigSelector = &metav1.LabelSelector{
+				MatchExpressions: []metav1.LabelSelectorRequirement{
+					{
+						Key:      "machineconfiguration.openshift.io/role",
+						Operator: metav1.LabelSelectorOpIn,
+						Values:   []string{depnodes.RoleWorker, roleMCPTest},
+					},
+				},
+			}
+			testMCP.Spec.NodeSelector = &metav1.LabelSelector{
+				MatchLabels: map[string]string{getLabelRoleMCPTest(): ""},
+			}
+
+			Expect(fxt.Client.Create(context.TODO(), testMCP)).To(Succeed())
+			defer func() {
+				By(fmt.Sprintf("CLEANUP: deleting mcp: %q", testMCP.Name))
+				Expect(fxt.Client.Delete(context.TODO(), testMCP)).To(Succeed())
+
+				err := wait.With(fxt.Client).
+					Interval(configuration.MachineConfigPoolUpdateInterval).
+					Timeout(configuration.MachineConfigPoolUpdateTimeout).
+					ForMachineConfigPoolDeleted(context.TODO(), testMCP)
+				Expect(err).ToNot(HaveOccurred())
+			}()
+			// keep the mcp with zero machine count intentionally to avoid reboots
+
+			testNG := nropv1.NodeGroup{
+				MachineConfigPoolSelector: &metav1.LabelSelector{
+					MatchLabels: testMCP.Labels,
+				},
+			}
+
+			By(fmt.Sprintf("modifying the NUMAResourcesOperator nodeGroups field to include new group: %q labels %q", testMCP.Name, testMCP.Labels))
+			newLogLevel := operatorv1.Trace
+			Eventually(func(g Gomega) {
+				// we need that for the current ResourceVersion
+				g.Expect(fxt.Client.Get(ctx, client.ObjectKeyFromObject(initialNroOperObj), &nroOperObj)).To(Succeed())
+
+				newNGs := append(nroOperObj.Spec.NodeGroups, testNG)
+				nroOperObj.Spec.NodeGroups = newNGs
+				nroOperObj.Spec.ExporterImage = serialconfig.GetRteCiImage()
+
+				if nroOperObj.Spec.LogLevel == operatorv1.Trace {
+					newLogLevel = operatorv1.Debug
+				}
+				nroOperObj.Spec.LogLevel = newLogLevel
+
+				g.Expect(fxt.Client.Update(ctx, &nroOperObj)).To(Succeed())
+			}).WithTimeout(5 * time.Minute).WithPolling(30 * time.Second).Should(Succeed())
+
+			defer func() {
+				By("CLEANUP: reverting the changes under the NUMAResourcesOperator object")
+				var nroOperObj nropv1.NUMAResourcesOperator
+				Eventually(func(g Gomega) {
+					// we need that for the current ResourceVersion
+					g.Expect(fxt.Client.Get(ctx, client.ObjectKeyFromObject(initialNroOperObj), &nroOperObj)).To(Succeed())
+
+					nroOperObj.Spec = initialNroOperObj.Spec
+					g.Expect(fxt.Client.Update(ctx, &nroOperObj)).To(Succeed())
+				}).WithTimeout(10 * time.Minute).WithPolling(30 * time.Second).Should(Succeed())
+			}() //end of defer
+
+			By("verify new RTE daemonset is created and all the RTE dameonsets are updated")
+			updatedNro := nropv1.NUMAResourcesOperator{}
+			Eventually(func(g Gomega) {
+				Expect(fxt.Client.Get(ctx, client.ObjectKeyFromObject(initialNroOperObj), &updatedNro)).To(Succeed())
+				dss, err := objects.GetDaemonSetsOwnedBy(fxt.Client, updatedNro.ObjectMeta)
+				g.Expect(err).ToNot(HaveOccurred())
+				// assumption 1:1 mapping to testMCP
+				g.Expect(dss).To(HaveLen(len(updatedNro.Spec.NodeGroups)), "daemonsets found owned by NRO object doesn't align with specified NodeGroups")
+
+				for _, ds := range dss {
+					By(fmt.Sprintf("check RTE daemonset %q", ds.Name))
+					if ds.Name == objectnames.GetComponentName(updatedNro.Name, roleMCPTest) {
+						By("check the correct match labels for the new RTE daemonset")
+						g.Expect(ds.Spec.Template.Spec.NodeSelector).To(Equal(testMCP.Spec.NodeSelector.MatchLabels))
+					}
+					By("check the correct image")
+					cnt := ds.Spec.Template.Spec.Containers[0]
+					g.Expect(cnt.Image).To(Equal(serialconfig.GetRteCiImage()))
+
+					By("checking the correct LogLevel")
+					found, match := matchLogLevelToKlog(&cnt, newLogLevel)
+					g.Expect(found).To(BeTrue(), "-v flag doesn't exist in container %q args under DaemonSet: %q", cnt.Name, ds.Name)
+					g.Expect(match).To(BeTrue(), "LogLevel %s doesn't match the existing -v flag in container: %q managed by DaemonSet: %q", updatedNro.Spec.LogLevel, cnt.Name, ds.Name)
+				}
+			}).WithTimeout(10*time.Minute).WithPolling(30*time.Second).Should(Succeed(), "failed to update RTE daemonset node selector")
+		})
+
 		It("[test_id:54916][tier2][schedrst] should be able to modify the configurable values under the NUMAResourcesScheduler CR", Label("tier2", "schedrst"), Label("feature:schedrst"), func() {
 			initialNroSchedObj := &nropv1.NUMAResourcesScheduler{}
 			nroSchedKey := objects.NROSchedObjectKey()