ceph: update CephNFS to use ".nfs" pool in newer ceph versions

This commit updates the CephNFS CR to make the RADOS settings optional
for Ceph versions above 16.2.7 due to the NFS module changes in Ceph.
The changes in Ceph make it so the RADOS pool is always ".nfs" and the
RADOS namespace is always the name of the NFS cluster.

This commit also handles the changes in Ceph Pacific versions before 16.2.7
where the default pool name is "nfs-ganesha" instead of ".nfs".

Closes: #8450
Signed-off-by: Joseph Sawaya <jsawaya@redhat.com>

Documentation/ceph-nfs-crd.md

@@ -25,6 +25,8 @@ metadata:
   name: my-nfs
   namespace: rook-ceph
 spec:
+  # rados property is not used in versions of Ceph equal to or greater than 
+  # 16.2.7, see note in RADOS settings section below.
   rados:
     # RADOS pool where NFS client recovery data and per-daemon configs are
     # stored. In this example the data pool for the "myfs" filesystem is used.
@@ -91,6 +93,8 @@ ceph dashboard set-ganesha-clusters-rados-pool-namespace <cluster_id>:<pool_name
 * `pool`: The pool where ganesha recovery backend and supplemental configuration objects will be stored
 * `namespace`: The namespace in `pool` where ganesha recovery backend and supplemental configuration objects will be stored
 
+> **NOTE**: The RADOS settings aren't used in Ceph versions equal to or greater than Pacific 16.2.7, default values are used instead ".nfs" for the RADOS pool and the CephNFS CR's name for the RADOS namespace. However, RADOS settings are mandatory for versions preceding Pacific 16.2.7.
+
 > **NOTE**: Don't use EC pools for NFS because ganesha uses omap in the recovery objects and grace db. EC pools do not support omap.
 
 ## EXPORT Block Configuration

cluster/charts/rook-ceph/templates/resources.yaml

@@ -5656,6 +5656,7 @@ spec:
               properties:
                 rados:
                   description: RADOS is the Ganesha RADOS specification
+                  nullable: true
                   properties:
                     namespace:
                       description: Namespace is the RADOS namespace where NFS client recovery data is stored.
@@ -6258,7 +6259,6 @@ spec:
                     - active
                   type: object
               required:
-                - rados
                 - server
               type: object
             status:

cluster/examples/kubernetes/ceph/crds.yaml

@@ -5653,6 +5653,7 @@ spec:
               properties:
                 rados:
                   description: RADOS is the Ganesha RADOS specification
+                  nullable: true
                   properties:
                     namespace:
                       description: Namespace is the RADOS namespace where NFS client recovery data is stored.
@@ -6255,7 +6256,6 @@ spec:
                     - active
                   type: object
               required:
-                - rados
                 - server
               type: object
             status:

cluster/examples/kubernetes/ceph/nfs-test.yaml

@@ -4,6 +4,7 @@ metadata:
   name: my-nfs
   namespace: rook-ceph # namespace:cluster
 spec:
+  # rados settings aren't necessary in Ceph Versions equal to or greater than Pacific 16.2.7
   rados:
     # RADOS pool where NFS client recovery data is stored.
     # In this example the data pool for the "myfs" filesystem is used.

pkg/apis/ceph.rook.io/v1/types.go

@@ -1571,7 +1571,9 @@ type CephNFSList struct {
 // NFSGaneshaSpec represents the spec of an nfs ganesha server
 type NFSGaneshaSpec struct {
 	// RADOS is the Ganesha RADOS specification
-	RADOS GaneshaRADOSSpec `json:"rados"`
+	// +nullable
+	// +optional
+	RADOS GaneshaRADOSSpec `json:"rados,omitempty"`
 
 	// Server is the Ganesha Server specification
 	Server GaneshaServerSpec `json:"server"`

pkg/operator/ceph/nfs/controller.go

@@ -31,6 +31,7 @@ import (
 	opconfig "github.com/rook/rook/pkg/operator/ceph/config"
 	opcontroller "github.com/rook/rook/pkg/operator/ceph/controller"
 	"github.com/rook/rook/pkg/operator/ceph/reporting"
+	"github.com/rook/rook/pkg/operator/ceph/version"
 	"github.com/rook/rook/pkg/operator/k8sutil"
 	appsv1 "k8s.io/api/apps/v1"
 	v1 "k8s.io/api/core/v1"
@@ -50,6 +51,9 @@ const (
 	controllerName = "ceph-nfs-controller"
 )
 
+// Version of Ceph where NFS default pool name changes to ".nfs"
+var cephNFSChangeVersion = version.CephVersion{Major: 16, Minor: 2, Extra: 7}
+
 var logger = capnslog.NewPackageLogger("github.com/rook/rook", controllerName)
 
 // List of object resources to watch by the controller
@@ -198,6 +202,20 @@ func (r *ReconcileCephNFS) reconcile(request reconcile.Request) (reconcile.Resul
 	}
 	r.clusterInfo.CephVersion = currentCephVersion
 
+	// Octopus: Customization is allowed, so don't change the pool and namespace
+	// Pacific before 16.2.7: No customization, default pool name is nfs-ganesha
+	// Pacific after 16.2.7: No customization, default pool name is .nfs
+	// This code is changes the pool and namespace to the correct values if the version is Pacific.
+	// If the version precedes Pacific it doesn't change it at all and the values used are what the user provided in the spec.
+	if r.clusterInfo.CephVersion.IsAtLeastPacific() {
+		if r.clusterInfo.CephVersion.IsAtLeast(cephNFSChangeVersion) {
+			cephNFS.Spec.RADOS.Pool = postNFSChangeDefaultPoolName
+		} else {
+			cephNFS.Spec.RADOS.Pool = preNFSChangeDefaultPoolName
+		}
+		cephNFS.Spec.RADOS.Namespace = cephNFS.Name
+	}
+
 	// DELETE: the CR was deleted
 	if !cephNFS.GetDeletionTimestamp().IsZero() {
 		logger.Infof("deleting ceph nfs %q", cephNFS.Name)
@@ -220,6 +238,9 @@ func (r *ReconcileCephNFS) reconcile(request reconcile.Request) (reconcile.Resul
 	if err := validateGanesha(r.context, r.clusterInfo, cephNFS); err != nil {
 		return reconcile.Result{}, errors.Wrapf(err, "invalid ceph nfs %q arguments", cephNFS.Name)
 	}
+	if err := fetchOrCreatePool(r.context, r.clusterInfo, cephNFS); err != nil {
+		return reconcile.Result{}, errors.Wrap(err, "failed to fetch or create RADOS pool")
+	}
 
 	// CREATE/UPDATE
 	logger.Debug("reconciling ceph nfs deployments")

pkg/operator/ceph/nfs/controller_test.go

@@ -28,6 +28,7 @@ import (
 	rookclient "github.com/rook/rook/pkg/client/clientset/versioned/fake"
 	"github.com/rook/rook/pkg/client/clientset/versioned/scheme"
 	"github.com/rook/rook/pkg/clusterd"
+	"github.com/rook/rook/pkg/operator/ceph/cluster/mon"
 	cephver "github.com/rook/rook/pkg/operator/ceph/version"
 	"github.com/rook/rook/pkg/operator/k8sutil"
 	"github.com/rook/rook/pkg/operator/test"
@@ -278,3 +279,131 @@ func TestGetGaneshaConfigObject(t *testing.T) {
 	logger.Infof("Config Object for Nautilus is %s", res)
 	assert.Equal(t, "conf-my-nfs.a", res)
 }
+
+func TestFetchOrCreatePool(t *testing.T) {
+	ctx := context.TODO()
+	cephNFS := &cephv1.CephNFS{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      name,
+			Namespace: namespace,
+		},
+		Spec: cephv1.NFSGaneshaSpec{
+			Server: cephv1.GaneshaServerSpec{
+				Active: 1,
+			},
+		},
+		TypeMeta: controllerTypeMeta,
+	}
+	executor := &exectest.MockExecutor{
+		MockExecuteCommandWithOutput: func(command string, args ...string) (string, error) {
+			return "", nil
+		},
+	}
+	clientset := test.New(t, 3)
+	c := &clusterd.Context{
+		Executor:      executor,
+		RookClientset: rookclient.NewSimpleClientset(),
+		Clientset:     clientset,
+	}
+	// Mock clusterInfo
+	secrets := map[string][]byte{
+		"fsid":         []byte(name),
+		"mon-secret":   []byte("monsecret"),
+		"admin-secret": []byte("adminsecret"),
+	}
+	secret := &v1.Secret{
+		ObjectMeta: metav1.ObjectMeta{
+			Name:      "rook-ceph-mon",
+			Namespace: namespace,
+		},
+		Data: secrets,
+		Type: k8sutil.RookType,
+	}
+	_, err := c.Clientset.CoreV1().Secrets(namespace).Create(ctx, secret, metav1.CreateOptions{})
+	assert.NoError(t, err)
+	clusterInfo, _, _, err := mon.LoadClusterInfo(c, namespace)
+	if err != nil {
+		return
+	}
+
+	err = fetchOrCreatePool(c, clusterInfo, cephNFS)
+	assert.NoError(t, err)
+
+	executor = &exectest.MockExecutor{
+		MockExecuteCommandWithOutput: func(command string, args ...string) (string, error) {
+			if args[1] == "pool" && args[2] == "get" {
+				return "Error", errors.New("failed to get pool")
+			}
+			return "", nil
+		},
+	}
+
+	c.Executor = executor
+	err = fetchOrCreatePool(c, clusterInfo, cephNFS)
+	assert.Error(t, err)
+
+	executor = &exectest.MockExecutor{
+		MockExecuteCommandWithOutput: func(command string, args ...string) (string, error) {
+			if args[1] == "pool" && args[2] == "get" {
+				return "Error", errors.New("failed to get pool: unrecognized pool")
+			}
+			return "", nil
+		},
+	}
+
+	c.Executor = executor
+	err = fetchOrCreatePool(c, clusterInfo, cephNFS)
+	assert.Error(t, err)
+
+	clusterInfo.CephVersion = cephver.CephVersion{
+		Major: 16,
+		Minor: 2,
+		Extra: 6,
+	}
+
+	executor = &exectest.MockExecutor{
+		MockExecuteCommandWithOutput: func(command string, args ...string) (string, error) {
+			if args[1] == "pool" && args[2] == "get" {
+				return "Error", errors.New("failed to get pool: unrecognized pool")
+			}
+			return "", nil
+		},
+	}
+
+	c.Executor = executor
+	err = fetchOrCreatePool(c, clusterInfo, cephNFS)
+	assert.NoError(t, err)
+
+	executor = &exectest.MockExecutor{
+		MockExecuteCommandWithOutput: func(command string, args ...string) (string, error) {
+			if args[1] == "pool" && args[2] == "get" {
+				return "Error", errors.New("failed to get pool: unrecognized pool")
+			}
+			if args[1] == "pool" && args[2] == "create" {
+				return "Error", errors.New("creating pool failed")
+			}
+			return "", nil
+		},
+	}
+
+	c.Executor = executor
+	err = fetchOrCreatePool(c, clusterInfo, cephNFS)
+	assert.Error(t, err)
+
+	executor = &exectest.MockExecutor{
+		MockExecuteCommandWithOutput: func(command string, args ...string) (string, error) {
+			if args[1] == "pool" && args[2] == "get" {
+				return "Error", errors.New("unrecognized pool")
+			}
+			if args[1] == "pool" && args[2] == "application" {
+				return "Error", errors.New("enabling pool failed")
+			}
+			return "", nil
+		},
+	}
+
+	c.Executor = executor
+	err = fetchOrCreatePool(c, clusterInfo, cephNFS)
+	assert.Error(t, err)
+
+}

pkg/operator/ceph/nfs/nfs.go

@@ -20,6 +20,7 @@ package nfs
 import (
 	"context"
 	"fmt"
+	"strings"
 
 	"github.com/banzaicloud/k8s-objectmatcher/patch"
 	"github.com/pkg/errors"
@@ -37,6 +38,10 @@ import (
 
 const (
 	ganeshaRadosGraceCmd = "ganesha-rados-grace"
+	// Default RADOS pool name after the NFS changes in Ceph
+	postNFSChangeDefaultPoolName = ".nfs"
+	// Default RADOS pool name before the NFS changes in Ceph
+	preNFSChangeDefaultPoolName = "nfs-ganesha"
 )
 
 var updateDeploymentAndWait = opmon.UpdateCephDeploymentAndWait
@@ -268,16 +273,45 @@ func validateGanesha(context *clusterd.Context, clusterInfo *cephclient.ClusterI
 		return errors.New("missing RADOS.pool")
 	}
 
+	if n.Spec.RADOS.Namespace == "" {
+		return errors.New("missing RADOS.namespace")
+	}
+
 	// Ganesha server properties
 	if n.Spec.Server.Active == 0 {
 		return errors.New("at least one active server required")
 	}
 
+	return nil
+}
+
+// create and enable default RADOS pool
+func createDefaultNFSRADOSPool(context *clusterd.Context, clusterInfo *cephclient.ClusterInfo, defaultRadosPoolName string) error {
+	args := []string{"osd", "pool", "create", defaultRadosPoolName}
+	_, err := cephclient.NewCephCommand(context, clusterInfo, args).Run()
+	if err != nil {
+		return err
+	}
+	args = []string{"osd", "pool", "application", "enable", defaultRadosPoolName, "nfs"}
+	_, err = cephclient.NewCephCommand(context, clusterInfo, args).Run()
+	if err != nil {
+		return err
+	}
+	return nil
+}
+
+func fetchOrCreatePool(context *clusterd.Context, clusterInfo *cephclient.ClusterInfo, n *cephv1.CephNFS) error {
 	// The existence of the pool provided in n.Spec.RADOS.Pool is necessary otherwise addRADOSConfigFile() will fail
 	_, err := cephclient.GetPoolDetails(context, clusterInfo, n.Spec.RADOS.Pool)
 	if err != nil {
+		if strings.Contains(err.Error(), "unrecognized pool") && clusterInfo.CephVersion.IsAtLeastPacific() {
+			err := createDefaultNFSRADOSPool(context, clusterInfo, n.Spec.RADOS.Pool)
+			if err != nil {
+				return errors.Wrapf(err, "failed to find %q pool and unable to create it", n.Spec.RADOS.Pool)
+			}
+			return nil
+		}
 		return errors.Wrapf(err, "pool %q not found", n.Spec.RADOS.Pool)
 	}
-
 	return nil
 }