mirrors/ceph-csi
1
0
Fork 0
mirror of https://github.com/ceph/ceph-csi.git synced 2024-12-18 02:50:30 +00:00
Code Issues Packages Projects Releases Wiki Activity

Remove mount cache for cephfs

Browse Source 
PR #282 introduces the mount cache to
solve cephfs fuse mount issue when cephfs plugin pod
restarts .This is not working as intended. This PR removes
the code for maintainability.
Signed-off-by: Madhu Rajanna <madhupr007@gmail.com>
This commit is contained in:
Madhu Rajanna 2020-02-10 14:21:05 +05:30 committed by mergify[bot]
parent 669dc4536f
commit 034b123478
8 changed files with 21 additions and 420 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

5
charts/ceph-csi-cephfs/templates/nodeplugin-daemonset.yaml
View File

@ -85,7 +85,6 @@ spec:
- "--v=5"
- "--drivername=$(DRIVER_NAME)"
- "--metadatastorage=k8s_configmap"
- "--mountcachedir=/mount-cache-dir"
env:
- name: POD_IP
valueFrom:
@ -109,8 +108,6 @@ spec:
add: ["SYS_ADMIN"]
allowPrivilegeEscalation: true
volumeMounts:
- name: mount-cache-dir
mountPath: /mount-cache-dir
- name: socket-dir
mountPath: /csi
- name: mountpoint-dir
@ -161,8 +158,6 @@ spec:
{{ toYaml .Values.nodeplugin.plugin.resources | indent 12 }}
{{- end }}
volumes:
- name: mount-cache-dir
emptyDir: {}
- name: socket-dir
hostPath:
path: {{ .Values.socketDir }}

4
cmd/cephcsi.go
View File

@ -60,6 +60,7 @@ func init() {
flag.BoolVar(&conf.IsNodeServer, "nodeserver", false, "start cephcsi node server")
// cephfs related flags
// marking this as deprecated, remove it in next major release
flag.StringVar(&conf.MountCacheDir, "mountcachedir", "", "mount info cache save dir")
flag.BoolVar(&conf.ForceKernelCephFS, "forcecephkernelclient", false, "enable Ceph Kernel clients on kernel < 4.17 which support quotas")
@ -172,6 +173,9 @@ func main() {
driver.Run(&conf, cp)
case cephfsType:
if conf.MountCacheDir != "" {
klog.Warning("mountcachedir option is deprecated")
}
driver := cephfs.NewDriver()
driver.Run(&conf, cp)

5
deploy/cephfs/kubernetes/csi-cephfsplugin.yaml
View File

@ -63,7 +63,6 @@ spec:
- "--v=5"
- "--drivername=cephfs.csi.ceph.com"
- "--metadatastorage=k8s_configmap"
- "--mountcachedir=/mount-cache-dir"
- "--metricsport=8090"
- "--metricspath=/metrics"
- "--enablegrpcmetrics=false"
@ -84,8 +83,6 @@ spec:
value: unix:///csi/csi.sock
imagePullPolicy: "IfNotPresent"
volumeMounts:
- name: mount-cache-dir
mountPath: /mount-cache-dir
- name: socket-dir
mountPath: /csi
- name: mountpoint-dir
@ -130,8 +127,6 @@ spec:
mountPath: /csi
imagePullPolicy: "IfNotPresent"
volumes:
- name: mount-cache-dir
emptyDir: {}
- name: socket-dir
hostPath:
path: /var/lib/kubelet/plugins/cephfs.csi.ceph.com/

3
docs/deploy-cephfs.md
View File

@ -44,12 +44,11 @@ that should be resolved in v14.2.3.
**Available command line arguments:**
| Option | Default value | Description |
| -------------------------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| ------------------------- | --------------------------- | ------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------ |
| `--endpoint` | `unix://tmp/csi.sock` | CSI endpoint, must be a UNIX socket |
| `--drivername` | `cephfs.csi.ceph.com` | Name of the driver (Kubernetes: `provisioner` field in StorageClass must correspond to this value) |
| `--nodeid` | _empty_ | This node's ID |
| `--type` | _empty_ | Driver type `[rbd | cephfs]` If the driver type is set to `rbd` it will act as a `rbd plugin` or if it's set to `cephfs` will act as a `cephfs plugin` |
| `--mountcachedir` | _empty_ | Volume mount cache info save dir. If left unspecified, the dirver will not record mount info, or it will save mount info and when driver restart it will remount volume it cached. |
| `--instanceid` | "default" | Unique ID distinguishing this instance of Ceph CSI among other instances, when sharing Ceph clusters across CSI instances for provisioning |
| `--pluginpath` | "/var/lib/kubelet/plugins/" | The location of cephcsi plugin on host |
| `--metadatastorage` | _empty_ | Points to where older (1.0.0 or older plugin versions) metadata about provisioned volumes are kept, as file or in as k8s configmap (`node` or `k8s_configmap` respectively) |

8
pkg/cephfs/driver.go
View File

@ -115,14 +115,6 @@ func (fs *Driver) Run(conf *util.Config, cachePersister util.CachePersister) {
// Update namespace for storing keys into a specific namespace on RADOS, in the CephFS
// metadata pool
volJournal.SetNamespace(radosNamespace)
initVolumeMountCache(conf.DriverName, conf.MountCacheDir)
if conf.MountCacheDir != "" {
if err := remountCachedVolumes(); err != nil {
klog.Warningf("failed to remount cached volumes: %v", err)
// ignore remount fail
}
}
// Initialize default library driver
fs.cd = csicommon.NewCSIDriver(conf.DriverName, util.DriverVersion, conf.NodeID)

330
pkg/cephfs/mountcache.go
View File

@ -1,330 +0,0 @@
/*
Copyright 2019 The Ceph-CSI 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 cephfs
import (
"context"
"encoding/base64"
"os"
"sync"
"syscall"
"time"
"github.com/ceph/ceph-csi/pkg/util"
"github.com/pkg/errors"
"k8s.io/klog"
)
type volumeMountCacheEntry struct {
DriverVersion string `json:"driverVersion"`
VolumeID string `json:"volumeID"`
Mounter string `json:"mounter"`
Secrets map[string]string `json:"secrets"`
StagingPath string `json:"stagingPath"`
TargetPaths map[string]bool `json:"targetPaths"`
CreateTime time.Time `json:"createTime"`
}
type volumeMountCacheMap struct {
volumes map[string]volumeMountCacheEntry
nodeCacheStore util.NodeCache
}
var (
volumeMountCachePrefix = "cephfs-mount-cache-"
volumeMountCache volumeMountCacheMap
volumeMountCacheMtx sync.Mutex
)
func initVolumeMountCache(driverName, mountCacheDir string) {
volumeMountCache.volumes = make(map[string]volumeMountCacheEntry)
volumeMountCache.nodeCacheStore.BasePath = mountCacheDir
volumeMountCache.nodeCacheStore.CacheDir = driverName
klog.Infof("mount-cache: name: %s, version: %s, mountCacheDir: %s", driverName, util.DriverVersion, mountCacheDir)
}
func remountCachedVolumes() error {
if err := util.CreateMountPoint(volumeMountCache.nodeCacheStore.BasePath); err != nil {
klog.Errorf("mount-cache: failed to create %s: %v", volumeMountCache.nodeCacheStore.BasePath, err)
return err
}
var remountFailCount, remountSuccCount int64
me := &volumeMountCacheEntry{}
err := volumeMountCache.nodeCacheStore.ForAll(volumeMountCachePrefix, me, func(identifier string) error {
volID := me.VolumeID
if volOpts, vid, err := newVolumeOptionsFromVolID(context.TODO(), me.VolumeID, nil, decodeCredentials(me.Secrets)); err != nil {
if err, ok := err.(util.ErrKeyNotFound); ok {
klog.Infof("mount-cache: image key not found, assuming the volume %s to be already deleted (%v)", volID, err)
if err := volumeMountCache.nodeCacheStore.Delete(genVolumeMountCacheFileName(volID)); err == nil {
klog.Infof("mount-cache: metadata not found, delete volume cache entry for volume %s", volID)
}
}
} else {
// update Mounter from mount cache
volOpts.Mounter = me.Mounter
if err := mountOneCacheEntry(volOpts, vid, me); err == nil {
remountSuccCount++
volumeMountCache.volumes[me.VolumeID] = *me
klog.Infof("mount-cache: successfully remounted volume %s", volID)
} else {
remountFailCount++
klog.Errorf("mount-cache: failed to remount volume %s", volID)
}
}
return nil
})
if err != nil {
klog.Infof("mount-cache: metastore list cache fail %v", err)
return err
}
if remountFailCount > 0 {
klog.Infof("mount-cache: successfully remounted %d volumes, failed to remount %d volumes", remountSuccCount, remountFailCount)
} else {
klog.Infof("mount-cache: successfully remounted %d volumes", remountSuccCount)
}
return nil
}
func mountOneCacheEntry(volOptions *volumeOptions, vid *volumeIdentifier, me *volumeMountCacheEntry) error {
volumeMountCacheMtx.Lock()
defer volumeMountCacheMtx.Unlock()
var (
err error
cr *util.Credentials
)
volID := vid.VolumeID
if volOptions.ProvisionVolume {
cr, err = util.NewAdminCredentials(decodeCredentials(me.Secrets))
if err != nil {
return err
}
defer cr.DeleteCredentials()
volOptions.RootPath, err = getVolumeRootPathCeph(context.TODO(), volOptions, cr, volumeID(vid.FsSubvolName))
if err != nil {
return err
}
} else {
cr, err = util.NewUserCredentials(decodeCredentials(me.Secrets))
if err != nil {
return err
}
defer cr.DeleteCredentials()
}
err = cleanupMountPoint(me.StagingPath)
if err != nil {
klog.Infof("mount-cache: failed to cleanup volume mount point %s, remove it: %s %v", volID, me.StagingPath, err)
return err
}
isMnt, err := util.IsMountPoint(me.StagingPath)
if err != nil {
isMnt = false
klog.Infof("mount-cache: failed to check volume mounted %s: %s %v", volID, me.StagingPath, err)
}
if !isMnt {
m, err := newMounter(volOptions)
if err != nil {
klog.Errorf("mount-cache: failed to create mounter for volume %s: %v", volID, err)
return err
}
if err := m.mount(context.TODO(), me.StagingPath, cr, volOptions); err != nil {
klog.Errorf("mount-cache: failed to mount volume %s: %v", volID, err)
return err
}
}
mountOptions := []string{"bind"}
for targetPath, readOnly := range me.TargetPaths {
if err := cleanupMountPoint(targetPath); err == nil {
if err := bindMount(context.TODO(), me.StagingPath, targetPath, readOnly, mountOptions); err != nil {
klog.Errorf("mount-cache: failed to bind-mount volume %s: %s %s %v %v",
volID, me.StagingPath, targetPath, readOnly, err)
} else {
klog.Infof("mount-cache: successfully bind-mounted volume %s: %s %s %v",
volID, me.StagingPath, targetPath, readOnly)
}
}
}
return nil
}
func cleanupMountPoint(mountPoint string) error {
if _, err := os.Stat(mountPoint); err != nil {
if isCorruptedMnt(err) {
klog.Infof("mount-cache: corrupted mount point %s, need unmount", mountPoint)
err := execCommandErr(context.TODO(), "umount", mountPoint)
if err != nil {
klog.Infof("mount-cache: failed to umount %s %v", mountPoint, err)
// ignore error return err
}
}
}
if _, err := os.Stat(mountPoint); err != nil {
klog.Errorf("mount-cache: failed to stat mount point %s %v", mountPoint, err)
return err
}
return nil
}
func isCorruptedMnt(err error) bool {
var underlyingError error
switch pe := err.(type) {
case nil:
return false
case *os.PathError:
underlyingError = pe.Err
case *os.LinkError:
underlyingError = pe.Err
case *os.SyscallError:
underlyingError = pe.Err
default:
return false
}
CorruptedErrors := []error{
syscall.ENOTCONN, syscall.ESTALE, syscall.EIO, syscall.EACCES}
for _, v := range CorruptedErrors {
if underlyingError == v {
return true
}
}
return false
}
func genVolumeMountCacheFileName(volID string) string {
cachePath := volumeMountCachePrefix + volID
return cachePath
}
func (mc *volumeMountCacheMap) isEnable() bool {
// if mount cache dir unset, disable state
return mc.nodeCacheStore.BasePath != ""
}
func (mc *volumeMountCacheMap) nodeStageVolume(ctx context.Context, volID, stagingTargetPath, mounter string, secrets map[string]string) error {
if !mc.isEnable() {
return nil
}
volumeMountCacheMtx.Lock()
defer volumeMountCacheMtx.Unlock()
lastTargetPaths := make(map[string]bool)
me, ok := volumeMountCache.volumes[volID]
if ok {
if me.StagingPath == stagingTargetPath {
klog.Warningf(util.Log(ctx, "mount-cache: node unexpected restage volume for volume %s"), volID)
return nil
}
lastTargetPaths = me.TargetPaths
klog.Warningf(util.Log(ctx, "mount-cache: node stage volume ignore last cache entry for volume %s"), volID)
}
me = volumeMountCacheEntry{DriverVersion: util.DriverVersion}
me.VolumeID = volID
me.Secrets = encodeCredentials(secrets)
me.StagingPath = stagingTargetPath
me.TargetPaths = lastTargetPaths
me.Mounter = mounter
me.CreateTime = time.Now()
volumeMountCache.volumes[volID] = me
return mc.nodeCacheStore.Create(genVolumeMountCacheFileName(volID), me)
}
func (mc *volumeMountCacheMap) nodeUnStageVolume(volID string) error {
if !mc.isEnable() {
return nil
}
volumeMountCacheMtx.Lock()
defer volumeMountCacheMtx.Unlock()
delete(volumeMountCache.volumes, volID)
return mc.nodeCacheStore.Delete(genVolumeMountCacheFileName(volID))
}
func (mc *volumeMountCacheMap) nodePublishVolume(ctx context.Context, volID, targetPath string, readOnly bool) error {
if !mc.isEnable() {
return nil
}
volumeMountCacheMtx.Lock()
defer volumeMountCacheMtx.Unlock()
_, ok := volumeMountCache.volumes[volID]
if !ok {
return errors.New("mount-cache: node publish volume failed to find cache entry for volume")
}
volumeMountCache.volumes[volID].TargetPaths[targetPath] = readOnly
return mc.updateNodeCache(ctx, volID)
}
func (mc *volumeMountCacheMap) nodeUnPublishVolume(ctx context.Context, volID, targetPath string) error {
if !mc.isEnable() {
return nil
}
volumeMountCacheMtx.Lock()
defer volumeMountCacheMtx.Unlock()
_, ok := volumeMountCache.volumes[volID]
if !ok {
return errors.New("mount-cache: node unpublish volume failed to find cache entry for volume")
}
delete(volumeMountCache.volumes[volID].TargetPaths, targetPath)
return mc.updateNodeCache(ctx, volID)
}
func (mc *volumeMountCacheMap) updateNodeCache(ctx context.Context, volID string) error {
me := volumeMountCache.volumes[volID]
if err := volumeMountCache.nodeCacheStore.Delete(genVolumeMountCacheFileName(volID)); err == nil {
klog.Infof(util.Log(ctx, "mount-cache: metadata not found, delete mount cache failed for volume %s"), volID)
}
return mc.nodeCacheStore.Create(genVolumeMountCacheFileName(volID), me)
}
func encodeCredentials(input map[string]string) (output map[string]string) {
output = make(map[string]string)
for key, value := range input {
nKey := base64.StdEncoding.EncodeToString([]byte(key))
nValue := base64.StdEncoding.EncodeToString([]byte(value))
output[nKey] = nValue
}
return output
}
func decodeCredentials(input map[string]string) (output map[string]string) {
output = make(map[string]string)
for key, value := range input {
nKey, err := base64.StdEncoding.DecodeString(key)
if err != nil {
klog.Errorf("mount-cache: decode secret fail")
continue
}
nValue, err := base64.StdEncoding.DecodeString(value)
if err != nil {
klog.Errorf("mount-cache: decode secret fail")
continue
}
output[string(nKey)] = string(nValue)
}
return output
}

38
pkg/cephfs/mountcache_test.go
View File

@ -1,38 +0,0 @@
package cephfs
import (
"testing"
)
func init() {
}
func TestMountOneCacheEntry(t *testing.T) {
}
func TestRemountHisMountedPath(t *testing.T) {
}
func TestNodeStageVolume(t *testing.T) {
}
func TestNodeUnStageVolume(t *testing.T) {
}
func TestNodePublishVolume(t *testing.T) {
}
func TestNodeUnpublishVolume(t *testing.T) {
}
func TestEncodeDecodeCredentials(t *testing.T) {
secrets := make(map[string]string)
secrets["user_1"] = "value_1"
enSecrets := encodeCredentials(secrets)
deSecrets := decodeCredentials(enSecrets)
for key, value := range secrets {
if deSecrets[key] != value {
t.Errorf("key %s of credentials's value %s change after decode %s ", key, value, deSecrets[key])
}
}
}

16
pkg/cephfs/nodeserver.go
View File

@ -154,9 +154,6 @@ func (*NodeServer) mount(ctx context.Context, volOptions *volumeOptions, req *cs
klog.Errorf(util.Log(ctx, "failed to mount volume %s: %v"), volID, err)
return status.Error(codes.Internal, err.Error())
}
if err := volumeMountCache.nodeStageVolume(ctx, req.GetVolumeId(), stagingTargetPath, volOptions.Mounter, req.GetSecrets()); err != nil {
klog.Warningf(util.Log(ctx, "mount-cache: failed to stage volume %s %s: %v"), volID, stagingTargetPath, err)
}
return nil
}
@ -209,10 +206,6 @@ func (ns *NodeServer) NodePublishVolume(ctx context.Context, req *csi.NodePublis
return nil, status.Error(codes.Internal, err.Error())
}
if err = volumeMountCache.nodePublishVolume(ctx, volID, targetPath, req.GetReadonly()); err != nil {
klog.Warningf(util.Log(ctx, "mount-cache: failed to publish volume %s %s: %v"), volID, targetPath, err)
}
klog.Infof(util.Log(ctx, "cephfs: successfully bind-mounted volume %s to %s"), volID, targetPath)
// #nosec - allow anyone to write inside the target path
@ -241,10 +234,6 @@ func (ns *NodeServer) NodeUnpublishVolume(ctx context.Context, req *csi.NodeUnpu
}
defer ns.VolumeLocks.Release(volID)
if err = volumeMountCache.nodeUnPublishVolume(ctx, volID, targetPath); err != nil {
klog.Warningf(util.Log(ctx, "mount-cache: failed to unpublish volume %s %s: %v"), volID, targetPath, err)
}
// Unmount the bind-mount
if err = unmountVolume(ctx, targetPath); err != nil {
return nil, status.Error(codes.Internal, err.Error())
@ -275,11 +264,6 @@ func (ns *NodeServer) NodeUnstageVolume(ctx context.Context, req *csi.NodeUnstag
defer ns.VolumeLocks.Release(volID)
stagingTargetPath := req.GetStagingTargetPath()
if err = volumeMountCache.nodeUnStageVolume(volID); err != nil {
klog.Warningf(util.Log(ctx, "mount-cache: failed to unstage volume %s %s: %v"), volID, stagingTargetPath, err)
}
// Unmount the volume
if err = unmountVolume(ctx, stagingTargetPath); err != nil {
return nil, status.Error(codes.Internal, err.Error())