Merge pull request #195 from nats-io/nats-v2-default

Update to use NATS v2 image by default

pkg/cluster/cluster.go

@@ -42,6 +42,7 @@ import (
 	"github.com/nats-io/nats-operator/pkg/debug"
 	kubernetesutil "github.com/nats-io/nats-operator/pkg/util/kubernetes"
 	stringutil "github.com/nats-io/nats-operator/pkg/util/strings"
+	"github.com/nats-io/nats-operator/pkg/util/versionCheck"
 )
 
 var (
@@ -354,8 +355,8 @@ func (c *Cluster) createPod() (*v1.Pod, error) {
 // It does this by trying to make the "gnatsd" process enter the "lame duck" mode before actually attempting to delete the pod.
 // This is done in a best-effort basis, since the NATS version running in the pod may not support this mode.
 // For that reason, we just log any errors without actually failing and proceed to the actual deletion of the pod.
-func (c *Cluster) tryGracefulPodDeletion(pod *v1.Pod) error {
-	if err := c.enterLameDuckModeAndWaitTermination(pod); err != nil {
+func (c *Cluster) tryGracefulPodDeletion(pod *v1.Pod, version string) error {
+	if err := c.enterLameDuckModeAndWaitTermination(pod, version); err != nil {
 		c.logger.Warn(err)
 	}
 	return c.deletePod(pod)
@@ -522,13 +523,14 @@ func (c *Cluster) isDebugLoggerEnabled() bool {
 // enterLameDuckModeAndWaitTermination execs into the "nats" container of the specified pod and attempts to send the "ldm" signal to the "gnatsd" process.
 // In case this succeeds, the funcion blocks until the "nats" container reaches the "Terminated" state (indicating that the "lame duck" mode has been entered and NATS is ready to shutdown) or until a timeout is reached.
 // Otherwise, it returns an error which should be handled by the caller.
-func (c *Cluster) enterLameDuckModeAndWaitTermination(pod *v1.Pod) error {
-	// Try to place NATS in "lame duck" mode by sending the "gnatsd" process the "ldm" signal.
+func (c *Cluster) enterLameDuckModeAndWaitTermination(pod *v1.Pod, version string) error {
+	// Try to place NATS in "lame duck" mode by sending the process the "ldm" signal.
 	// We wait for at most "podExecTimeout" for the "exec" command to return a result.
 	ctx, fn := context.WithTimeout(context.Background(), podExecTimeout)
 	defer fn()
+
 	args := []string{
-		constants.NatsBinaryPath,
+		versionCheck.ServerBinaryPath(version),
 		"-sl",
 		fmt.Sprintf("ldm=%s", constants.PidFilePath),
 	}

pkg/cluster/reconcile.go

@@ -45,7 +45,7 @@ func (c *Cluster) reconcileSize() error {
 		// Remove extra pods as required in order to meet the desired size.
 		// As we remove each pod, we must update the config secret so that routes are re-computed.
 		for idx := currentSize - 1; idx >= desiredSize; idx-- {
-			if err := c.tryGracefulPodDeletion(pods[idx]); err != nil {
+			if err := c.tryGracefulPodDeletion(pods[idx], c.cluster.Spec.Version); err != nil {
 				return err
 			}
 			if err := c.updateConfigSecret(); err != nil {
@@ -94,7 +94,7 @@ func (c *Cluster) reconcileVersion() error {
 		for _, pod := range pods {
 			if kubernetesutil.GetNATSVersion(pod) != c.cluster.Spec.Version {
 				c.maybeUpgradeMgmtService()
-				return c.upgradePod(pod)
+				return c.upgradePod(pod, desiredVersion)
 			}
 		}
 	}

pkg/cluster/upgrade.go

@@ -29,8 +29,8 @@ import (
 // In order to do that, we first try to make NATS enter the "lame duck" mode.
 // If we succeed, we adopt a special upgrade procedure since the pod (or at least its "nats" container) will have been terminated and can't be upgraded directly.
 // If we fail, we stick to the usual method of upgrading the container's "image" field to the desired version.
-func (c *Cluster) upgradePod(pod *v1.Pod) error {
-	if err := c.enterLameDuckModeAndWaitTermination(pod); err != nil {
+func (c *Cluster) upgradePod(pod *v1.Pod, version string) error {
+	if err := c.enterLameDuckModeAndWaitTermination(pod, version); err != nil {
 		c.logger.Warn(err)
 		return c.upgradeRunningPod(pod)
 	}

pkg/constants/constants.go

@@ -16,7 +16,7 @@ package constants
 
 const (
 	// DefaultNatsVersion is the nats server version to use.
-	DefaultNatsVersion = "1.4.0"
+	DefaultNatsVersion = "2.0.0"
 
 	// ClientPort is the port for the clients.
 	ClientPort = 4222
@@ -58,7 +58,7 @@ const (
 	PidFileVolumeName = "pid"
 
 	// PidFileName is the pid file name.
-	PidFileName = "gnatsd.pid"
+	PidFileName = "nats.pid"
 
 	// PidFileMountPath is the absolute path to the directory where NATS
 	// will be leaving its pid file.
@@ -128,8 +128,6 @@ const (
 	DefaultBootconfigImage         = "connecteverything/nats-boot-config"
 	DefaultBootconfigImageTag      = "0.5.2"
 
-	// NatsBinaryPath is the path to the NATS binary inside the main container.
-	NatsBinaryPath = "/gnatsd"
 	// NatsContainerName is the name of the main container.
 	NatsContainerName = "nats"
 

pkg/util/kubernetes/kubernetes.go

@@ -45,6 +45,7 @@ import (
 	"github.com/nats-io/nats-operator/pkg/conf"
 	"github.com/nats-io/nats-operator/pkg/constants"
 	"github.com/nats-io/nats-operator/pkg/util/retryutil"
+	"github.com/nats-io/nats-operator/pkg/util/versionCheck"
 )
 
 const (
@@ -920,7 +921,7 @@ func NewNatsPodSpec(namespace, name, clusterName string, cs v1alpha2.ClusterSpec
 	// Rely on the shared configuration map for configuring the cluster.
 	retries := strconv.Itoa(constants.ConnectRetries)
 	cmd := []string{
-		constants.NatsBinaryPath,
+		versionCheck.ServerBinaryPath(cs.Version),
 		"-c",
 		constants.ConfigFilePath,
 		"-P",

pkg/util/versionCheck/versionCheck.go

@@ -0,0 +1,29 @@
+package versionCheck
+
+import (
+	"strconv"
+	"strings"
+)
+
+const (
+	// OldNatsBinaryPath is the path to the NATS binary inside the
+	// main container, before NATS Server v2.
+	OldNatsBinaryPath = "/gnatsd"
+
+	// NatsBinaryPath after v2 release.
+	NatsBinaryPath = "/nats-server"
+)
+
+func ServerBinaryPath(version string) string {
+	v := strings.Split(version, ".")
+	if len(v) > 0 {
+		majorVersion, err := strconv.Atoi(v[0])
+		if err != nil {
+			return NatsBinaryPath
+		}
+		if majorVersion < 2 {
+			return OldNatsBinaryPath
+		}
+	}
+	return NatsBinaryPath
+}

pkg/util/versionCheck/versionCheck_test.go

@@ -0,0 +1,37 @@
+package versionCheck
+
+import (
+	"testing"
+)
+
+func TestServerBinaryPath(t *testing.T) {
+	got := ServerBinaryPath("2.0.0")
+	expected := NatsBinaryPath
+	if got != expected {
+		t.Fatalf("Expected %q, got: %q", expected, got)
+	}
+
+	got = ServerBinaryPath("2.1.8")
+	expected = NatsBinaryPath
+	if got != expected {
+		t.Fatalf("Expected %q, got: %q", expected, got)
+	}
+
+	got = ServerBinaryPath("2.1.8-RC18")
+	expected = NatsBinaryPath
+	if got != expected {
+		t.Fatalf("Expected %q, got: %q", expected, got)
+	}
+
+	got = ServerBinaryPath("0.1.8")
+	expected = OldNatsBinaryPath
+	if got != expected {
+		t.Fatalf("Expected %q, got: %q", expected, got)
+	}
+
+	got = ServerBinaryPath("1.4.1")
+	expected = OldNatsBinaryPath
+	if got != expected {
+		t.Fatalf("Expected %q, got: %q", expected, got)
+	}
+}

test/e2e/basic_test.go

@@ -60,6 +60,37 @@ func TestCreateCluster(t *testing.T) {
 	}
 }
 
+// TestCreateClusterV2 creates a NatsCluster resource and waits for the full mesh to be formed.
+func TestCreateClusterV2(t *testing.T) {
+	var (
+		size    = 3
+		version = "2.0.0"
+	)
+
+	var (
+		natsCluster *natsv1alpha2.NatsCluster
+		err         error
+	)
+
+	// Create a NatsCluster resource with three members.
+	if natsCluster, err = f.CreateCluster(f.Namespace, "test-nats-", size, version); err != nil {
+		t.Fatal(err)
+	}
+	// Make sure we cleanup the NatsCluster resource after we're done testing.
+	defer func() {
+		if err = f.DeleteCluster(natsCluster); err != nil {
+			t.Error(err)
+		}
+	}()
+
+	// Wait until the full mesh is formed.
+	ctx, fn := context.WithTimeout(context.Background(), waitTimeout)
+	defer fn()
+	if err = f.WaitUntilFullMeshWithVersion(ctx, natsCluster, size, version); err != nil {
+		t.Fatal(err)
+	}
+}
+
 // TestPauseControl creates a NatsCluster resource and waits for the full mesh to be formed.
 // Then, it pauses control of the NatsCluster resource and scales it up to five nodes, expecting the operation to NOT be performed.
 // Finally, it resumes control of the NatsCluster resource and waits for the full five-node mesh to be formed.

test/e2e/lame_duck_mode_test.go

@@ -42,10 +42,8 @@ func TestLameDuckModeWhenScalingDown(t *testing.T) {
 	var (
 		initialSize = 3
 		finalSize   = 1
-		// TODO Replace with an adequate stable tag once there is one.
-		version = "5d86964"
-		// TODO Remove once the "nats" image has an adequate stable tag.
-		serverImage = "natsop2018/gnatsd"
+		serverImage = "synadia/nats-server"
+		version     = "2.0.0"
 	)
 
 	var (

test/e2e/super_cluster_test.go

@@ -32,15 +32,13 @@ import (
 func TestCreateServerWithGateways(t *testing.T) {
 	var (
 		size    = 1
-		image   = "synadia/nats-server"
-		version = "edge-v2.0.0-RC12"
+		version = "2.0.0"
 		nc      *natsv1alpha2.NatsCluster
 		err     error
 	)
 	nc, err = f.CreateCluster(f.Namespace, "", size, version,
 		func(cluster *natsv1alpha2.NatsCluster) {
 			cluster.Name = "test-nats-gw"
-			cluster.Spec.ServerImage = image
 
 			cluster.Spec.ServerConfig = &natsv1alpha2.ServerConfig{
 				Debug: true,
@@ -138,7 +136,7 @@ func TestCreateServerWithGatewayAndLeafnodes(t *testing.T) {
 	var (
 		size    = 1
 		image   = "synadia/nats-server"
-		version = "edge-v2.0.0-RC12"
+		version = "2.0.0"
 		nc      *natsv1alpha2.NatsCluster
 		err     error
 

test/e2e/tls_test.go

@@ -144,8 +144,7 @@ func TestCreateClusterWithVerifyAndMap(t *testing.T) {
 		// The NatsCluster resource must be called "nats" in
 		// order for the pre-provisioned certificates to work.
 		natsCluster.Name = "nats-verify"
-		natsCluster.Spec.ServerImage = "wallyqs/nats-server"
-		natsCluster.Spec.Version = "edge-2.0.0-RC5"
+		natsCluster.Spec.Version = "2.0.0"
 
 		// Enable TLS using pre-provisioned certificates.
 		natsCluster.Spec.TLS = &natsv1alpha2.TLSConfig{

version/version.go

@@ -15,6 +15,6 @@
 package version
 
 var (
-	OperatorVersion = "0.5.0-v1alpha2+git"
+	OperatorVersion = "0.6.0-v1alpha2+git"
 	GitSHA          = "Not provided"
 )