ea89d62710
Automatic merge from submit-queue (batch tested with PRs 49081, 49318, 49219, 48989, 48486) Use MetricsStatsFs to expose RBD volume plugin metrics. **What this PR does / why we need it**: We need to monitor RBD volume usage of our cluster and configure alerts if RBD volume is nearly full. Users of cluster also need to see usage history graph on Grafana. This PR use MetricsStatsFs to implement MetricsProvider interface of RBD plugin (same as `gce_pd`), so kubelet /stat/summary can expose RBD volume stats. **Special notes for your reviewer**: cc @rootfs **Release note**: ```release-note NONE ```
533 lines
16 KiB
Go
533 lines
16 KiB
Go
/*
|
|
Copyright 2014 The Kubernetes Authors.
|
|
|
|
Licensed under the Apache License, Version 2.0 (the "License");
|
|
you may not use this file except in compliance with the License.
|
|
You may obtain a copy of the License at
|
|
|
|
http://www.apache.org/licenses/LICENSE-2.0
|
|
|
|
Unless required by applicable law or agreed to in writing, software
|
|
distributed under the License is distributed on an "AS IS" BASIS,
|
|
WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.
|
|
See the License for the specific language governing permissions and
|
|
limitations under the License.
|
|
*/
|
|
|
|
package rbd
|
|
|
|
import (
|
|
"fmt"
|
|
dstrings "strings"
|
|
|
|
"github.com/golang/glog"
|
|
"k8s.io/api/core/v1"
|
|
"k8s.io/apimachinery/pkg/api/resource"
|
|
metav1 "k8s.io/apimachinery/pkg/apis/meta/v1"
|
|
"k8s.io/apimachinery/pkg/types"
|
|
"k8s.io/apimachinery/pkg/util/sets"
|
|
"k8s.io/apimachinery/pkg/util/uuid"
|
|
clientset "k8s.io/client-go/kubernetes"
|
|
"k8s.io/kubernetes/pkg/util/mount"
|
|
"k8s.io/kubernetes/pkg/util/strings"
|
|
"k8s.io/kubernetes/pkg/volume"
|
|
volutil "k8s.io/kubernetes/pkg/volume/util"
|
|
"k8s.io/kubernetes/pkg/volume/util/volumehelper"
|
|
"k8s.io/utils/exec"
|
|
)
|
|
|
|
var (
|
|
supportedFeatures = sets.NewString("layering")
|
|
)
|
|
|
|
// This is the primary entrypoint for volume plugins.
|
|
func ProbeVolumePlugins() []volume.VolumePlugin {
|
|
return []volume.VolumePlugin{&rbdPlugin{nil, exec.New()}}
|
|
}
|
|
|
|
type rbdPlugin struct {
|
|
host volume.VolumeHost
|
|
exe exec.Interface
|
|
}
|
|
|
|
var _ volume.VolumePlugin = &rbdPlugin{}
|
|
var _ volume.PersistentVolumePlugin = &rbdPlugin{}
|
|
var _ volume.DeletableVolumePlugin = &rbdPlugin{}
|
|
var _ volume.ProvisionableVolumePlugin = &rbdPlugin{}
|
|
|
|
const (
|
|
rbdPluginName = "kubernetes.io/rbd"
|
|
secretKeyName = "key" // key name used in secret
|
|
rbdImageFormat1 = "1"
|
|
rbdImageFormat2 = "2"
|
|
rbdDefaultAdminId = "admin"
|
|
rbdDefaultAdminSecretNamespace = "default"
|
|
rbdDefaultPool = "rbd"
|
|
rbdDefaultUserId = rbdDefaultAdminId
|
|
)
|
|
|
|
func getPath(uid types.UID, volName string, host volume.VolumeHost) string {
|
|
return host.GetPodVolumeDir(uid, strings.EscapeQualifiedNameForDisk(rbdPluginName), volName)
|
|
}
|
|
|
|
func (plugin *rbdPlugin) Init(host volume.VolumeHost) error {
|
|
plugin.host = host
|
|
return nil
|
|
}
|
|
|
|
func (plugin *rbdPlugin) GetPluginName() string {
|
|
return rbdPluginName
|
|
}
|
|
|
|
func (plugin *rbdPlugin) GetVolumeName(spec *volume.Spec) (string, error) {
|
|
volumeSource, _, err := getVolumeSource(spec)
|
|
if err != nil {
|
|
return "", err
|
|
}
|
|
|
|
return fmt.Sprintf(
|
|
"%v:%v",
|
|
volumeSource.CephMonitors,
|
|
volumeSource.RBDImage), nil
|
|
}
|
|
|
|
func (plugin *rbdPlugin) CanSupport(spec *volume.Spec) bool {
|
|
if (spec.Volume != nil && spec.Volume.RBD == nil) || (spec.PersistentVolume != nil && spec.PersistentVolume.Spec.RBD == nil) {
|
|
return false
|
|
}
|
|
|
|
return true
|
|
}
|
|
|
|
func (plugin *rbdPlugin) RequiresRemount() bool {
|
|
return false
|
|
}
|
|
|
|
func (plugin *rbdPlugin) SupportsMountOption() bool {
|
|
return true
|
|
}
|
|
|
|
func (plugin *rbdPlugin) SupportsBulkVolumeVerification() bool {
|
|
return false
|
|
}
|
|
|
|
func (plugin *rbdPlugin) GetAccessModes() []v1.PersistentVolumeAccessMode {
|
|
return []v1.PersistentVolumeAccessMode{
|
|
v1.ReadWriteOnce,
|
|
v1.ReadOnlyMany,
|
|
}
|
|
}
|
|
|
|
func (plugin *rbdPlugin) NewMounter(spec *volume.Spec, pod *v1.Pod, _ volume.VolumeOptions) (volume.Mounter, error) {
|
|
var secret string
|
|
var err error
|
|
source, _ := plugin.getRBDVolumeSource(spec)
|
|
|
|
if source.SecretRef != nil {
|
|
if secret, err = parsePodSecret(pod, source.SecretRef.Name, plugin.host.GetKubeClient()); err != nil {
|
|
glog.Errorf("Couldn't get secret from %v/%v", pod.Namespace, source.SecretRef)
|
|
return nil, err
|
|
}
|
|
}
|
|
|
|
// Inject real implementations here, test through the internal function.
|
|
return plugin.newMounterInternal(spec, pod.UID, &RBDUtil{}, plugin.host.GetMounter(), secret)
|
|
}
|
|
|
|
func (plugin *rbdPlugin) getRBDVolumeSource(spec *volume.Spec) (*v1.RBDVolumeSource, bool) {
|
|
// rbd volumes used directly in a pod have a ReadOnly flag set by the pod author.
|
|
// rbd volumes used as a PersistentVolume gets the ReadOnly flag indirectly through the persistent-claim volume used to mount the PV
|
|
if spec.Volume != nil && spec.Volume.RBD != nil {
|
|
return spec.Volume.RBD, spec.Volume.RBD.ReadOnly
|
|
} else {
|
|
return spec.PersistentVolume.Spec.RBD, spec.ReadOnly
|
|
}
|
|
}
|
|
|
|
func (plugin *rbdPlugin) newMounterInternal(spec *volume.Spec, podUID types.UID, manager diskManager, mounter mount.Interface, secret string) (volume.Mounter, error) {
|
|
source, readOnly := plugin.getRBDVolumeSource(spec)
|
|
pool := source.RBDPool
|
|
id := source.RadosUser
|
|
keyring := source.Keyring
|
|
|
|
return &rbdMounter{
|
|
rbd: &rbd{
|
|
podUID: podUID,
|
|
volName: spec.Name(),
|
|
Image: source.RBDImage,
|
|
Pool: pool,
|
|
ReadOnly: readOnly,
|
|
manager: manager,
|
|
mounter: &mount.SafeFormatAndMount{Interface: mounter, Runner: exec.New()},
|
|
plugin: plugin,
|
|
MetricsProvider: volume.NewMetricsStatFS(getPath(podUID, spec.Name(), plugin.host)),
|
|
},
|
|
Mon: source.CephMonitors,
|
|
Id: id,
|
|
Keyring: keyring,
|
|
Secret: secret,
|
|
fsType: source.FSType,
|
|
mountOptions: volume.MountOptionFromSpec(spec),
|
|
}, nil
|
|
}
|
|
|
|
func (plugin *rbdPlugin) NewUnmounter(volName string, podUID types.UID) (volume.Unmounter, error) {
|
|
// Inject real implementations here, test through the internal function.
|
|
return plugin.newUnmounterInternal(volName, podUID, &RBDUtil{}, plugin.host.GetMounter())
|
|
}
|
|
|
|
func (plugin *rbdPlugin) newUnmounterInternal(volName string, podUID types.UID, manager diskManager, mounter mount.Interface) (volume.Unmounter, error) {
|
|
return &rbdUnmounter{
|
|
rbdMounter: &rbdMounter{
|
|
rbd: &rbd{
|
|
podUID: podUID,
|
|
volName: volName,
|
|
manager: manager,
|
|
mounter: &mount.SafeFormatAndMount{Interface: mounter, Runner: exec.New()},
|
|
plugin: plugin,
|
|
MetricsProvider: volume.NewMetricsStatFS(getPath(podUID, volName, plugin.host)),
|
|
},
|
|
Mon: make([]string, 0),
|
|
},
|
|
}, nil
|
|
}
|
|
|
|
func (plugin *rbdPlugin) ConstructVolumeSpec(volumeName, mountPath string) (*volume.Spec, error) {
|
|
rbdVolume := &v1.Volume{
|
|
Name: volumeName,
|
|
VolumeSource: v1.VolumeSource{
|
|
RBD: &v1.RBDVolumeSource{
|
|
CephMonitors: []string{},
|
|
},
|
|
},
|
|
}
|
|
return volume.NewSpecFromVolume(rbdVolume), nil
|
|
}
|
|
|
|
func (plugin *rbdPlugin) NewDeleter(spec *volume.Spec) (volume.Deleter, error) {
|
|
if spec.PersistentVolume != nil && spec.PersistentVolume.Spec.RBD == nil {
|
|
return nil, fmt.Errorf("spec.PersistentVolumeSource.Spec.RBD is nil")
|
|
}
|
|
class, err := volutil.GetClassForVolume(plugin.host.GetKubeClient(), spec.PersistentVolume)
|
|
if err != nil {
|
|
return nil, err
|
|
}
|
|
adminSecretName := ""
|
|
adminSecretNamespace := rbdDefaultAdminSecretNamespace
|
|
admin := ""
|
|
|
|
for k, v := range class.Parameters {
|
|
switch dstrings.ToLower(k) {
|
|
case "adminid":
|
|
admin = v
|
|
case "adminsecretname":
|
|
adminSecretName = v
|
|
case "adminsecretnamespace":
|
|
adminSecretNamespace = v
|
|
}
|
|
}
|
|
|
|
if admin == "" {
|
|
admin = rbdDefaultAdminId
|
|
}
|
|
secret, err := parsePVSecret(adminSecretNamespace, adminSecretName, plugin.host.GetKubeClient())
|
|
if err != nil {
|
|
return nil, fmt.Errorf("failed to get admin secret from [%q/%q]: %v", adminSecretNamespace, adminSecretName, err)
|
|
}
|
|
return plugin.newDeleterInternal(spec, admin, secret, &RBDUtil{})
|
|
}
|
|
|
|
func (plugin *rbdPlugin) newDeleterInternal(spec *volume.Spec, admin, secret string, manager diskManager) (volume.Deleter, error) {
|
|
return &rbdVolumeDeleter{
|
|
rbdMounter: &rbdMounter{
|
|
rbd: &rbd{
|
|
volName: spec.Name(),
|
|
Image: spec.PersistentVolume.Spec.RBD.RBDImage,
|
|
Pool: spec.PersistentVolume.Spec.RBD.RBDPool,
|
|
manager: manager,
|
|
plugin: plugin,
|
|
},
|
|
Mon: spec.PersistentVolume.Spec.RBD.CephMonitors,
|
|
adminId: admin,
|
|
adminSecret: secret,
|
|
}}, nil
|
|
}
|
|
|
|
func (plugin *rbdPlugin) NewProvisioner(options volume.VolumeOptions) (volume.Provisioner, error) {
|
|
return plugin.newProvisionerInternal(options, &RBDUtil{})
|
|
}
|
|
|
|
func (plugin *rbdPlugin) newProvisionerInternal(options volume.VolumeOptions, manager diskManager) (volume.Provisioner, error) {
|
|
return &rbdVolumeProvisioner{
|
|
rbdMounter: &rbdMounter{
|
|
rbd: &rbd{
|
|
manager: manager,
|
|
plugin: plugin,
|
|
},
|
|
},
|
|
options: options,
|
|
}, nil
|
|
}
|
|
|
|
type rbdVolumeProvisioner struct {
|
|
*rbdMounter
|
|
options volume.VolumeOptions
|
|
}
|
|
|
|
func (r *rbdVolumeProvisioner) Provision() (*v1.PersistentVolume, error) {
|
|
if !volume.AccessModesContainedInAll(r.plugin.GetAccessModes(), r.options.PVC.Spec.AccessModes) {
|
|
return nil, fmt.Errorf("invalid AccessModes %v: only AccessModes %v are supported", r.options.PVC.Spec.AccessModes, r.plugin.GetAccessModes())
|
|
}
|
|
|
|
if r.options.PVC.Spec.Selector != nil {
|
|
return nil, fmt.Errorf("claim Selector is not supported")
|
|
}
|
|
var err error
|
|
adminSecretName := ""
|
|
adminSecretNamespace := rbdDefaultAdminSecretNamespace
|
|
secretName := ""
|
|
secret := ""
|
|
imageFormat := rbdImageFormat1
|
|
fstype := ""
|
|
|
|
for k, v := range r.options.Parameters {
|
|
switch dstrings.ToLower(k) {
|
|
case "monitors":
|
|
arr := dstrings.Split(v, ",")
|
|
for _, m := range arr {
|
|
r.Mon = append(r.Mon, m)
|
|
}
|
|
case "adminid":
|
|
r.adminId = v
|
|
case "adminsecretname":
|
|
adminSecretName = v
|
|
case "adminsecretnamespace":
|
|
adminSecretNamespace = v
|
|
case "userid":
|
|
r.Id = v
|
|
case "pool":
|
|
r.Pool = v
|
|
case "usersecretname":
|
|
secretName = v
|
|
case "imageformat":
|
|
imageFormat = v
|
|
case "imagefeatures":
|
|
arr := dstrings.Split(v, ",")
|
|
for _, f := range arr {
|
|
if !supportedFeatures.Has(f) {
|
|
return nil, fmt.Errorf("invalid feature %q for volume plugin %s, supported features are: %v", f, r.plugin.GetPluginName(), supportedFeatures)
|
|
} else {
|
|
r.imageFeatures = append(r.imageFeatures, f)
|
|
}
|
|
}
|
|
case volume.VolumeParameterFSType:
|
|
fstype = v
|
|
default:
|
|
return nil, fmt.Errorf("invalid option %q for volume plugin %s", k, r.plugin.GetPluginName())
|
|
}
|
|
}
|
|
// sanity check
|
|
if imageFormat != rbdImageFormat1 && imageFormat != rbdImageFormat2 {
|
|
return nil, fmt.Errorf("invalid ceph imageformat %s, expecting %s or %s",
|
|
imageFormat, rbdImageFormat1, rbdImageFormat2)
|
|
}
|
|
r.imageFormat = imageFormat
|
|
if adminSecretName == "" {
|
|
return nil, fmt.Errorf("missing Ceph admin secret name")
|
|
}
|
|
if secret, err = parsePVSecret(adminSecretNamespace, adminSecretName, r.plugin.host.GetKubeClient()); err != nil {
|
|
return nil, fmt.Errorf("failed to get admin secret from [%q/%q]: %v", adminSecretNamespace, adminSecretName, err)
|
|
}
|
|
r.adminSecret = secret
|
|
if len(r.Mon) < 1 {
|
|
return nil, fmt.Errorf("missing Ceph monitors")
|
|
}
|
|
if secretName == "" {
|
|
return nil, fmt.Errorf("missing user secret name")
|
|
}
|
|
if r.adminId == "" {
|
|
r.adminId = rbdDefaultAdminId
|
|
}
|
|
if r.Pool == "" {
|
|
r.Pool = rbdDefaultPool
|
|
}
|
|
if r.Id == "" {
|
|
r.Id = r.adminId
|
|
}
|
|
|
|
// create random image name
|
|
image := fmt.Sprintf("kubernetes-dynamic-pvc-%s", uuid.NewUUID())
|
|
r.rbdMounter.Image = image
|
|
rbd, sizeMB, err := r.manager.CreateImage(r)
|
|
if err != nil {
|
|
glog.Errorf("rbd: create volume failed, err: %v", err)
|
|
return nil, err
|
|
}
|
|
glog.Infof("successfully created rbd image %q", image)
|
|
pv := new(v1.PersistentVolume)
|
|
metav1.SetMetaDataAnnotation(&pv.ObjectMeta, volumehelper.VolumeDynamicallyCreatedByKey, "rbd-dynamic-provisioner")
|
|
rbd.SecretRef = new(v1.LocalObjectReference)
|
|
rbd.SecretRef.Name = secretName
|
|
rbd.RadosUser = r.Id
|
|
rbd.FSType = fstype
|
|
pv.Spec.PersistentVolumeSource.RBD = rbd
|
|
pv.Spec.PersistentVolumeReclaimPolicy = r.options.PersistentVolumeReclaimPolicy
|
|
pv.Spec.AccessModes = r.options.PVC.Spec.AccessModes
|
|
if len(pv.Spec.AccessModes) == 0 {
|
|
pv.Spec.AccessModes = r.plugin.GetAccessModes()
|
|
}
|
|
pv.Spec.Capacity = v1.ResourceList{
|
|
v1.ResourceName(v1.ResourceStorage): resource.MustParse(fmt.Sprintf("%dMi", sizeMB)),
|
|
}
|
|
return pv, nil
|
|
}
|
|
|
|
type rbdVolumeDeleter struct {
|
|
*rbdMounter
|
|
}
|
|
|
|
func (r *rbdVolumeDeleter) GetPath() string {
|
|
return getPath(r.podUID, r.volName, r.plugin.host)
|
|
}
|
|
|
|
func (r *rbdVolumeDeleter) Delete() error {
|
|
return r.manager.DeleteImage(r)
|
|
}
|
|
|
|
type rbd struct {
|
|
volName string
|
|
podUID types.UID
|
|
Pool string
|
|
Image string
|
|
ReadOnly bool
|
|
plugin *rbdPlugin
|
|
mounter *mount.SafeFormatAndMount
|
|
// Utility interface that provides API calls to the provider to attach/detach disks.
|
|
manager diskManager
|
|
volume.MetricsProvider
|
|
}
|
|
|
|
func (rbd *rbd) GetPath() string {
|
|
// safe to use PodVolumeDir now: volume teardown occurs before pod is cleaned up
|
|
return getPath(rbd.podUID, rbd.volName, rbd.plugin.host)
|
|
}
|
|
|
|
type rbdMounter struct {
|
|
*rbd
|
|
// capitalized so they can be exported in persistRBD()
|
|
Mon []string
|
|
Id string
|
|
Keyring string
|
|
Secret string
|
|
fsType string
|
|
adminSecret string
|
|
adminId string
|
|
mountOptions []string
|
|
imageFormat string
|
|
imageFeatures []string
|
|
}
|
|
|
|
var _ volume.Mounter = &rbdMounter{}
|
|
|
|
func (b *rbd) GetAttributes() volume.Attributes {
|
|
return volume.Attributes{
|
|
ReadOnly: b.ReadOnly,
|
|
Managed: !b.ReadOnly,
|
|
SupportsSELinux: true,
|
|
}
|
|
}
|
|
|
|
// Checks prior to mount operations to verify that the required components (binaries, etc.)
|
|
// to mount the volume are available on the underlying node.
|
|
// If not, it returns an error
|
|
func (b *rbdMounter) CanMount() error {
|
|
return nil
|
|
}
|
|
|
|
func (b *rbdMounter) SetUp(fsGroup *int64) error {
|
|
return b.SetUpAt(b.GetPath(), fsGroup)
|
|
}
|
|
|
|
func (b *rbdMounter) SetUpAt(dir string, fsGroup *int64) error {
|
|
// diskSetUp checks mountpoints and prevent repeated calls
|
|
glog.V(4).Infof("rbd: attempting to SetUp and mount %s", dir)
|
|
err := diskSetUp(b.manager, *b, dir, b.mounter, fsGroup)
|
|
if err != nil {
|
|
glog.Errorf("rbd: failed to setup mount %s %v", dir, err)
|
|
}
|
|
return err
|
|
}
|
|
|
|
type rbdUnmounter struct {
|
|
*rbdMounter
|
|
}
|
|
|
|
var _ volume.Unmounter = &rbdUnmounter{}
|
|
|
|
// Unmounts the bind mount, and detaches the disk only if the disk
|
|
// resource was the last reference to that disk on the kubelet.
|
|
func (c *rbdUnmounter) TearDown() error {
|
|
return c.TearDownAt(c.GetPath())
|
|
}
|
|
|
|
func (c *rbdUnmounter) TearDownAt(dir string) error {
|
|
if pathExists, pathErr := volutil.PathExists(dir); pathErr != nil {
|
|
return fmt.Errorf("Error checking if path exists: %v", pathErr)
|
|
} else if !pathExists {
|
|
glog.Warningf("Warning: Unmount skipped because path does not exist: %v", dir)
|
|
return nil
|
|
}
|
|
return diskTearDown(c.manager, *c, dir, c.mounter)
|
|
}
|
|
|
|
func (plugin *rbdPlugin) execCommand(command string, args []string) ([]byte, error) {
|
|
cmd := plugin.exe.Command(command, args...)
|
|
return cmd.CombinedOutput()
|
|
}
|
|
|
|
func getVolumeSource(
|
|
spec *volume.Spec) (*v1.RBDVolumeSource, bool, error) {
|
|
if spec.Volume != nil && spec.Volume.RBD != nil {
|
|
return spec.Volume.RBD, spec.Volume.RBD.ReadOnly, nil
|
|
} else if spec.PersistentVolume != nil &&
|
|
spec.PersistentVolume.Spec.RBD != nil {
|
|
return spec.PersistentVolume.Spec.RBD, spec.ReadOnly, nil
|
|
}
|
|
|
|
return nil, false, fmt.Errorf("Spec does not reference a RBD volume type")
|
|
}
|
|
|
|
func parsePodSecret(pod *v1.Pod, secretName string, kubeClient clientset.Interface) (string, error) {
|
|
secret, err := volutil.GetSecretForPod(pod, secretName, kubeClient)
|
|
if err != nil {
|
|
glog.Errorf("failed to get secret from [%q/%q]", pod.Namespace, secretName)
|
|
return "", fmt.Errorf("failed to get secret from [%q/%q]", pod.Namespace, secretName)
|
|
}
|
|
return parseSecretMap(secret)
|
|
}
|
|
|
|
func parsePVSecret(namespace, secretName string, kubeClient clientset.Interface) (string, error) {
|
|
secret, err := volutil.GetSecretForPV(namespace, secretName, rbdPluginName, kubeClient)
|
|
if err != nil {
|
|
glog.Errorf("failed to get secret from [%q/%q]", namespace, secretName)
|
|
return "", fmt.Errorf("failed to get secret from [%q/%q]", namespace, secretName)
|
|
}
|
|
return parseSecretMap(secret)
|
|
}
|
|
|
|
// parseSecretMap locates the secret by key name.
|
|
func parseSecretMap(secretMap map[string]string) (string, error) {
|
|
if len(secretMap) == 0 {
|
|
return "", fmt.Errorf("empty secret map")
|
|
}
|
|
secret := ""
|
|
for k, v := range secretMap {
|
|
if k == secretKeyName {
|
|
return v, nil
|
|
}
|
|
secret = v
|
|
}
|
|
// If not found, the last secret in the map wins as done before
|
|
return secret, nil
|
|
}
|