e2e: block all master addresses

This way we can be sure that the kubelet can't communicate with the
master, even if falls-back to the internal/external IP (which seems to
be the case with DNS)

Issue #56787

test/e2e/apps/network_partition.go

@@ -198,10 +198,12 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 				By(fmt.Sprintf("Block traffic from node %s to the master", node.Name))
 				host, err := framework.GetNodeExternalIP(&node)
 				framework.ExpectNoError(err)
-				master := framework.GetMasterAddress(c)
+				masterAddresses := framework.GetAllMasterAddresses(c)
 				defer func() {
 					By(fmt.Sprintf("Unblock traffic from node %s to the master", node.Name))
-					framework.UnblockNetwork(host, master)
+					for _, masterAddress := range masterAddresses {
+						framework.UnblockNetwork(host, masterAddress)
+					}
 
 					if CurrentGinkgoTestDescription().Failed {
 						return
@@ -214,7 +216,9 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 					}
 				}()
 
-				framework.BlockNetwork(host, master)
+				for _, masterAddress := range masterAddresses {
+					framework.BlockNetwork(host, masterAddress)
+				}
 
 				By("Expect to observe node and pod status change from Ready to NotReady after network partition")
 				expectNodeReadiness(false, newNode)
@@ -576,10 +580,12 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 				By(fmt.Sprintf("Block traffic from node %s to the master", node.Name))
 				host, err := framework.GetNodeExternalIP(&node)
 				framework.ExpectNoError(err)
-				master := framework.GetMasterAddress(c)
+				masterAddresses := framework.GetAllMasterAddresses(c)
 				defer func() {
 					By(fmt.Sprintf("Unblock traffic from node %s to the master", node.Name))
-					framework.UnblockNetwork(host, master)
+					for _, masterAddress := range masterAddresses {
+						framework.UnblockNetwork(host, masterAddress)
+					}
 
 					if CurrentGinkgoTestDescription().Failed {
 						return
@@ -589,7 +595,9 @@ var _ = SIGDescribe("Network Partition [Disruptive] [Slow]", func() {
 					expectNodeReadiness(true, newNode)
 				}()
 
-				framework.BlockNetwork(host, master)
+				for _, masterAddress := range masterAddresses {
+					framework.BlockNetwork(host, masterAddress)
+				}
 
 				By("Expect to observe node and pod status change from Ready to NotReady after network partition")
 				expectNodeReadiness(false, newNode)

test/e2e/framework/networking_utils.go

@@ -952,7 +952,7 @@ func TestUnderTemporaryNetworkFailure(c clientset.Interface, ns string, node *v1
 	if err != nil {
 		Failf("Error getting node external ip : %v", err)
 	}
-	master := GetMasterAddress(c)
+	masterAddresses := GetAllMasterAddresses(c)
 	By(fmt.Sprintf("block network traffic from node %s to the master", node.Name))
 	defer func() {
 		// This code will execute even if setting the iptables rule failed.
@@ -960,14 +960,18 @@ func TestUnderTemporaryNetworkFailure(c clientset.Interface, ns string, node *v1
 		// had been inserted. (yes, we could look at the error code and ssh error
 		// separately, but I prefer to stay on the safe side).
 		By(fmt.Sprintf("Unblock network traffic from node %s to the master", node.Name))
-		UnblockNetwork(host, master)
+		for _, masterAddress := range masterAddresses {
+			UnblockNetwork(host, masterAddress)
+		}
 	}()
 
 	Logf("Waiting %v to ensure node %s is ready before beginning test...", resizeNodeReadyTimeout, node.Name)
 	if !WaitForNodeToBe(c, node.Name, v1.NodeReady, true, resizeNodeReadyTimeout) {
 		Failf("Node %s did not become ready within %v", node.Name, resizeNodeReadyTimeout)
 	}
-	BlockNetwork(host, master)
+	for _, masterAddress := range masterAddresses {
+		BlockNetwork(host, masterAddress)
+	}
 
 	Logf("Waiting %v for node %s to be not ready after simulated network failure", resizeNodeNotReadyTimeout, node.Name)
 	if !WaitForNodeToBe(c, node.Name, v1.NodeReady, false, resizeNodeNotReadyTimeout) {

test/e2e/framework/util.go

@@ -4934,19 +4934,28 @@ func getMaster(c clientset.Interface) Address {
 	return master
 }
 
-// GetMasterAddress returns the hostname/external IP/internal IP as appropriate for e2e tests on a particular provider
+// GetAllMasterAddresses returns all IP addresses on which the kubelet can reach the master.
-// which is the address of the interface used for communication with the kubelet.
+// It may return internal and external IPs, even if we expect for
-func GetMasterAddress(c clientset.Interface) string {
+// e.g. internal IPs to be used (issue #56787), so that we can be
+// sure to block the master fully during tests.
+func GetAllMasterAddresses(c clientset.Interface) []string {
 	master := getMaster(c)
+
+	var ips sets.String
 	switch TestContext.Provider {
 	case "gce", "gke":
-		return master.externalIP
+		if master.externalIP != "" {
+			ips.Insert(master.externalIP)
+		}
+		if master.internalIP != "" {
+			ips.Insert(master.internalIP)
+		}
 	case "aws":
-		return awsMasterIP
+		ips.Insert(awsMasterIP)
 	default:
 		Failf("This test is not supported for provider %s and should be disabled", TestContext.Provider)
 	}
-	return ""
+	return ips.List()
 }
 
 // GetNodeExternalIP returns node external IP concatenated with port 22 for ssh

test/e2e/network/firewall.go

@@ -173,16 +173,18 @@ var _ = SIGDescribe("Firewall rule", func() {
 		By("Checking well known ports on master and nodes are not exposed externally")
 		nodeAddrs := framework.NodeAddresses(nodes, v1.NodeExternalIP)
 		Expect(len(nodeAddrs)).NotTo(BeZero())
-		masterAddr := framework.GetMasterAddress(cs)
+		masterAddresses := framework.GetAllMasterAddresses(cs)
-		flag, _ := framework.TestNotReachableHTTPTimeout(masterAddr, ports.InsecureKubeControllerManagerPort, gce.FirewallTestTcpTimeout)
+		for _, masterAddr := range masterAddresses {
-		Expect(flag).To(BeTrue())
+			flag, _ := framework.TestNotReachableHTTPTimeout(masterAddr, ports.InsecureKubeControllerManagerPort, gce.FirewallTestTcpTimeout)
-		flag, _ = framework.TestNotReachableHTTPTimeout(masterAddr, ports.SchedulerPort, gce.FirewallTestTcpTimeout)
+			Expect(flag).To(BeTrue())
-		Expect(flag).To(BeTrue())
+			flag, _ = framework.TestNotReachableHTTPTimeout(masterAddr, ports.SchedulerPort, gce.FirewallTestTcpTimeout)
-		flag, _ = framework.TestNotReachableHTTPTimeout(nodeAddrs[0], ports.KubeletPort, gce.FirewallTestTcpTimeout)
+			Expect(flag).To(BeTrue())
-		Expect(flag).To(BeTrue())
+			flag, _ = framework.TestNotReachableHTTPTimeout(nodeAddrs[0], ports.KubeletPort, gce.FirewallTestTcpTimeout)
-		flag, _ = framework.TestNotReachableHTTPTimeout(nodeAddrs[0], ports.KubeletReadOnlyPort, gce.FirewallTestTcpTimeout)
+			Expect(flag).To(BeTrue())
-		Expect(flag).To(BeTrue())
+			flag, _ = framework.TestNotReachableHTTPTimeout(nodeAddrs[0], ports.KubeletReadOnlyPort, gce.FirewallTestTcpTimeout)
-		flag, _ = framework.TestNotReachableHTTPTimeout(nodeAddrs[0], ports.ProxyStatusPort, gce.FirewallTestTcpTimeout)
+			Expect(flag).To(BeTrue())
-		Expect(flag).To(BeTrue())
+			flag, _ = framework.TestNotReachableHTTPTimeout(nodeAddrs[0], ports.ProxyStatusPort, gce.FirewallTestTcpTimeout)
+			Expect(flag).To(BeTrue())
+		}
 	})
 })

test/e2e/scheduling/taint_based_evictions.go

@@ -128,12 +128,14 @@ var _ = SIGDescribe("TaintBasedEvictions [Serial]", func() {
 		// 	host, err = framework.GetNodeInternalIP(&node)
 		// }
 		framework.ExpectNoError(err)
-		master := framework.GetMasterAddress(cs)
+		masterAddresses := framework.GetAllMasterAddresses(cs)
 		taint := newUnreachableNoExecuteTaint()
 
 		defer func() {
 			By(fmt.Sprintf("Unblocking traffic from node %s to the master", node.Name))
-			framework.UnblockNetwork(host, master)
+			for _, masterAddress := range masterAddresses {
+				framework.UnblockNetwork(host, masterAddress)
+			}
 
 			if CurrentGinkgoTestDescription().Failed {
 				framework.Failf("Current e2e test has failed, so return from here.")
@@ -147,7 +149,9 @@ var _ = SIGDescribe("TaintBasedEvictions [Serial]", func() {
 			framework.ExpectNoError(err)
 		}()
 
-		framework.BlockNetwork(host, master)
+		for _, masterAddress := range masterAddresses {
+			framework.BlockNetwork(host, masterAddress)
+		}
 
 		By(fmt.Sprintf("Expecting to see node %q becomes NotReady", nodeName))
 		if !framework.WaitForNodeToBeNotReady(cs, nodeName, time.Minute*3) {