e2e: serial: ensure MCPs and nodes are healthy

At the setup of the test, check that MCPs and nodes are healthy enough
for the test to run by verifying that the MCPs are all `Updated` and non of the nodes is
`unschedulable`. We don't want to fail on any error so keep any
unexpected results at warning level.

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

Author: shajmakh

test/internal/fixture/fixture.go

@@ -36,6 +36,8 @@ import (
 	"sigs.k8s.io/controller-runtime/pkg/client"
 	"sigs.k8s.io/yaml"
 
+	machineconfigv1 "github.com/openshift/api/machineconfiguration/v1"
+
 	"github.com/k8stopologyawareschedwg/deployer/pkg/deployer"
 	nrtv1alpha2 "github.com/k8stopologyawareschedwg/noderesourcetopology-api/pkg/apis/topology/v1alpha2"
 
@@ -102,6 +104,7 @@ func SetupWithOptions(name string, nrtList nrtv1alpha2.NodeResourceTopologyList,
 	randomizeName := (options & OptionRandomizeName) == OptionRandomizeName
 	avoidCooldown := (options & OptionAvoidCooldown) == OptionAvoidCooldown
 	staticClusterData := (options & OptionStaticClusterData) == OptionStaticClusterData
+
 	ginkgo.By("set up the test namespace")
 	ns, err := setupNamespace(e2eclient.Client, name, randomizeName)
 	if err != nil {
@@ -133,6 +136,44 @@ func SetupWithOptions(name string, nrtList nrtv1alpha2.NodeResourceTopologyList,
 	}
 	klog.Infof("set up the fixture reference NRT List: %s", intnrt.ListToString(nrtList.Items, " fixture initial"))
 
+	ginkgo.By("warn about not updated MCPs")
+	var mcps machineconfigv1.MachineConfigPoolList
+	err = wait.PollUntilContextTimeout(ctx, 10*time.Second, 30*time.Second, true, func(ctx context.Context) (bool, error) {
+		err := e2eclient.Client.List(ctx, &mcps)
+		return err == nil, nil
+	})
+	if err != nil {
+		klog.Errorf("failed to pull MCP items: %v", err)
+	} else {
+		for _, mcp := range mcps.Items {
+			conditions := mcp.Status.Conditions
+			klog.InfoS("MCP status", "name", mcp.Name, "conditions", conditions)
+			for _, condition := range conditions {
+				if condition.Type == machineconfigv1.MachineConfigPoolUpdated {
+					if condition.Status != corev1.ConditionTrue {
+						klog.Warningf("MCP %q is not updated", mcp.Name)
+						break
+					}
+				}
+			}
+		}
+	}
+	ginkgo.By("warn about unschedulable nodes")
+	var nodes corev1.NodeList
+	err = wait.PollUntilContextTimeout(ctx, 10*time.Second, 30*time.Second, true, func(ctx context.Context) (bool, error) {
+		err := e2eclient.Client.List(ctx, &nodes)
+		return err == nil, nil
+	})
+	if err != nil {
+		klog.Errorf("failed to pull cluster nodes: %v", err)
+	} else {
+		for _, node := range nodes.Items {
+			if node.Spec.Unschedulable {
+				klog.Warningf("Node %q is unschedulable", node.Name)
+			}
+		}
+	}
+
 	return &Fixture{
 		Client:         e2eclient.Client,
 		K8sClient:      e2eclient.K8sClient,