MCO: update MCN condition transition test for clusters with no worker MCP nodes

Author: isabella-janssen

test/extended/machine_config/helpers.go

@@ -703,17 +703,38 @@ func WaitForNodeCurrentConfig(oc *exutil.CLI, nodeName string, config string) {
 	}, 5*time.Minute, 10*time.Second).Should(o.BeTrue(), "Timed out waiting for node '%v' to have a current config version of '%v'.", nodeName, config)
 }
 
-// `GetUpdatingNodeSNO` returns the SNO node when the `master` MCP of the cluster starts updating
-func GetUpdatingNodeSNO(oc *exutil.CLI, mcpName string) corev1.Node {
+// `GetCordonedNode` returns the cordoned node when the corresponding MCP starts updating
+func GetCordonedNode(oc *exutil.CLI, mcpName string) corev1.Node {
 	// Wait for the MCP to start updating
 	o.Expect(WaitForMCPConditionStatus(oc, mcpName, mcfgv1.MachineConfigPoolUpdating, corev1.ConditionTrue, 3*time.Minute, 2*time.Second)).NotTo(o.HaveOccurred(), "Waiting for 'Updating' status change failed.")
 
-	// SNO only has one node, so when the MCP is updating, the node is also updating
-	node, nodeErr := GetNodesByRole(oc, mcpName)
-	o.Expect(nodeErr).NotTo(o.HaveOccurred(), "Error getting nodes from %v MCP.", mcpName)
-	o.Expect(node).ShouldNot(o.BeEmpty(), "No nodes found for %v MCP.", mcpName)
+	// // SNO only has one node, so when the MCP is updating, the node is also updating
+	// node, nodeErr := GetNodesByRole(oc, mcpName)
+	// o.Expect(nodeErr).NotTo(o.HaveOccurred(), "Error getting nodes from %v MCP.", mcpName)
+	// o.Expect(node).ShouldNot(o.BeEmpty(), "No nodes found for %v MCP.", mcpName)
 
-	return node[0]
+	// Get first cordoned node & return it
+	var cordonedNode corev1.Node
+	o.Eventually(func() bool {
+		framework.Logf("Trying to get cordoned node in '%v' MCP.", mcpName)
+
+		// Get nodes in MCP
+		nodes, nodeErr := GetNodesByRole(oc, mcpName)
+		o.Expect(nodeErr).NotTo(o.HaveOccurred(), "Error getting nodes from %v MCP.", mcpName)
+		o.Expect(nodes).ShouldNot(o.BeEmpty(), "No nodes found for %v MCP.", mcpName)
+
+		// Loop through nodes to see which is cordoned
+		for _, node := range nodes {
+			if node.Spec.Unschedulable {
+				cordonedNode = node
+				return true
+			}
+		}
+
+		return false
+	}, 1*time.Minute, 5*time.Second).Should(o.BeTrue())
+
+	return cordonedNode
 }
 
 // `WaitForMCPConditionStatus` waits up to the desired timeout for the desired MCP condition to match the desired status (ex. wait until "Updating" is "True")

test/extended/machine_config/machine_config_node.go

@@ -6,6 +6,7 @@ import (
 	"fmt"
 	"os/exec"
 	"path/filepath"
+	"slices"
 	"time"
 
 	mcfgv1 "github.com/openshift/api/machineconfiguration/v1"
@@ -77,10 +78,19 @@ var _ = g.Describe("[sig-mco][OCPFeatureGate:MachineConfigNodes]", func() {
 		// regression tests handle the different conditions list.
 		SkipWhenFeatureGateEnabled(oc.AdminConfigClient(), "ImageModeStatusReporting")
 
-		if IsSingleNode(oc) {
-			ValidateMCNConditionTransitionsOnRebootlessUpdateSNO(oc, nodeDisruptionFixture, nodeDisruptionEmptyFixture, masterMCFixture)
-		} else {
-			ValidateMCNConditionTransitionsOnRebootlessUpdate(oc, nodeDisruptionFixture, nodeDisruptionEmptyFixture, customMCFixture, infraMCPFixture)
+		// Create client set for test
+		clientSet, clientErr := machineconfigclient.NewForConfig(oc.KubeFramework().ClientConfig())
+		o.Expect(clientErr).NotTo(o.HaveOccurred(), "Error creating client set for test.")
+
+		// Get MCPs to test for cluster
+		poolNames := GetRolesToTest(oc, clientSet)
+		framework.Logf("Validating MCN properties for node(s) in pool(s) '%v'.", poolNames)
+
+		// When the cluster has machines in the "worker" MCP, use a custom MCP to test the update
+		if slices.Contains(poolNames, worker) {
+			ValidateMCNConditionTransitionsOnRebootlessUpdate(oc, clientSet, nodeDisruptionFixture, nodeDisruptionEmptyFixture, customMCFixture, infraMCPFixture)
+		} else { // When there are no machines in the "worker" MCP, test the update by applying a MC targeting the "master" MCP
+			ValidateMCNConditionTransitionsOnRebootlessUpdateMaster(oc, clientSet, nodeDisruptionFixture, nodeDisruptionEmptyFixture, masterMCFixture)
 		}
 	})
 
@@ -168,17 +178,17 @@ func ValidateMCNPropertiesCustomMCP(oc *exutil.CLI, fixture string) {
 	o.Expect(mcnErr).NotTo(o.HaveOccurred(), fmt.Sprintf("Error validating MCN properties node in custom pool '%v'.", custom))
 }
 
-// `ValidateMCNConditionTransitions` checks that Conditions properly update on a node update
-// Note that a custom MCP is created for this test to limit the number of upgrading nodes &
-// decrease cleanup time.
-func ValidateMCNConditionTransitionsOnRebootlessUpdate(oc *exutil.CLI, nodeDisruptionFixture string, nodeDisruptionEmptyFixture string, mcFixture string, mcpFixture string) {
+// `ValidateMCNConditionTransitionsOnRebootlessUpdate` checks that the `Conditions` in an MCN
+// properly update on a node update in a custom MCP. The steps of this function are:
+//  1. Apply a node disruption policy
+//  2. Create a custom MCP with one node
+//  3. Apply a MC
+//  4. Validate the MCN conditions transition as expected throughout the update
+//  5. Clean up the test resources
+func ValidateMCNConditionTransitionsOnRebootlessUpdate(oc *exutil.CLI, clientSet *machineconfigclient.Clientset, nodeDisruptionFixture string, nodeDisruptionEmptyFixture string, mcFixture string, mcpFixture string) {
 	poolName := custom
 	mcName := fmt.Sprintf("90-%v-testfile", poolName)
 
-	// Create client set for test
-	clientSet, clientErr := machineconfigclient.NewForConfig(oc.KubeFramework().ClientConfig())
-	o.Expect(clientErr).NotTo(o.HaveOccurred(), "Error creating client set for test.")
-
 	// Grab a random worker node
 	workerNode := GetRandomNode(oc, worker)
 	o.Expect(workerNode.Name).NotTo(o.Equal(""), "Could not get a worker node.")
@@ -218,16 +228,17 @@ func ValidateMCNConditionTransitionsOnRebootlessUpdate(oc *exutil.CLI, nodeDisru
 	o.Expect(ConfirmUpdatedMCNStatus(clientSet, updatingNodeName)).Should(o.BeTrue(), "Error, all conditions must be 'False' when Updated=True.")
 }
 
-// `ValidateMCNConditionTransitionsSNO` checks that Conditions properly update on a node update
-// in Single Node Openshift
-func ValidateMCNConditionTransitionsOnRebootlessUpdateSNO(oc *exutil.CLI, nodeDisruptionFixture string, nodeDisruptionEmptyFixture string, mcFixture string) {
+// `ValidateMCNConditionTransitionsOnRebootlessUpdateMaster` checks that the `Conditions` in an MCN
+// properly update on a node update in the master MCP. The steps of this function are:
+//  1. Apply a node disruption policy
+//  2. Apply a MC
+//  3. Get the updating node
+//  4. Validate the MCN conditions transition as expected throughout the update
+//  5. Clean up the test resources
+func ValidateMCNConditionTransitionsOnRebootlessUpdateMaster(oc *exutil.CLI, clientSet *machineconfigclient.Clientset, nodeDisruptionFixture string, nodeDisruptionEmptyFixture string, mcFixture string) {
 	poolName := master
 	mcName := fmt.Sprintf("90-%v-testfile", poolName)
 
-	// Create client set for test
-	clientSet, clientErr := machineconfigclient.NewForConfig(oc.KubeFramework().ClientConfig())
-	o.Expect(clientErr).NotTo(o.HaveOccurred(), "Error creating client set for test.")
-
 	// Remove node disruption policy on test completion or failure
 	defer func() {
 		// Apply empty MachineConfiguration fixture to remove previously set NodeDisruptionPolicy
@@ -251,12 +262,13 @@ func ValidateMCNConditionTransitionsOnRebootlessUpdateSNO(oc *exutil.CLI, nodeDi
 		WaitForMCPToBeReady(oc, clientSet, poolName, 1)
 	}()
 
-	// Apply MC targeting worker node
+	// Apply MC targeting master MCP
 	mcErr := oc.Run("apply").Args("-f", mcFixture).Execute()
 	o.Expect(mcErr).NotTo(o.HaveOccurred(), "Could not apply MachineConfig.")
 
 	// Get the updating node
-	updatingNode := GetUpdatingNodeSNO(oc, poolName)
+	updatingNode := GetCordonedNode(oc, poolName)
+	o.Expect(updatingNode).NotTo(o.BeNil(), "Could not get updating node.")
 	framework.Logf("Node '%v' is updating.", updatingNode.Name)
 
 	// Validate transition through conditions for MCN