Commit 3fe8db86 authored by Kubernetes Submit Queue's avatar Kubernetes Submit Queue Committed by GitHub

Merge pull request #33270 from hpcloud/bug/33128

Automatic merge from submit-queue Don't rely on device name provided by Cinder See issue #33128 We can't rely on the device name provided by Cinder, and thus must perform detection based on the drive serial number (aka It's cinder ID) on the kubelet itself. This patch re-works the cinder volume attacher to ignore the supplied deviceName, and instead defer to the pre-existing GetDevicePath method to discover the device path based on it's serial number and /dev/disk/by-id mapping. This new behavior is controller by a config option, as falling back to the cinder value when we can't discover a device would risk devices not showing up, falling back to cinder's guess, and detecting the wrong disk as attached.
parents 356b28b7 ce8eda94
......@@ -86,11 +86,16 @@ type LoadBalancerOpts struct {
NodeSecurityGroupID string `gcfg:"node-security-group"`
}
type BlockStorageOpts struct {
TrustDevicePath bool `gcfg:"trust-device-path"` // See Issue #33128
}
// OpenStack is an implementation of cloud provider Interface for OpenStack.
type OpenStack struct {
provider *gophercloud.ProviderClient
region string
lbOpts LoadBalancerOpts
bsOpts BlockStorageOpts
// InstanceID of the server where this OpenStack object is instantiated.
localInstanceID string
}
......@@ -110,6 +115,7 @@ type Config struct {
Region string
}
LoadBalancer LoadBalancerOpts
BlockStorage BlockStorageOpts
}
func init() {
......@@ -146,6 +152,10 @@ func readConfig(config io.Reader) (Config, error) {
}
var cfg Config
// Set default values for config params
cfg.BlockStorage.TrustDevicePath = false
err := gcfg.ReadInto(&cfg, config)
return cfg, err
}
......@@ -200,6 +210,7 @@ func newOpenStack(cfg Config) (*OpenStack, error) {
provider: provider,
region: cfg.Global.Region,
lbOpts: cfg.LoadBalancer,
bsOpts: cfg.BlockStorage,
localInstanceID: id,
}
......
......@@ -76,6 +76,8 @@ func TestReadConfig(t *testing.T) {
monitor-delay = 1m
monitor-timeout = 30s
monitor-max-retries = 3
[BlockStorage]
trust-device-path = yes
`))
if err != nil {
t.Fatalf("Should succeed when a valid config is provided: %s", err)
......@@ -96,6 +98,9 @@ func TestReadConfig(t *testing.T) {
if cfg.LoadBalancer.MonitorMaxRetries != 3 {
t.Errorf("incorrect lb.monitormaxretries: %d", cfg.LoadBalancer.MonitorMaxRetries)
}
if cfg.BlockStorage.TrustDevicePath != true {
t.Errorf("incorrect bs.trustdevicepath: %v", cfg.BlockStorage.TrustDevicePath)
}
}
func TestToAuthOptions(t *testing.T) {
......@@ -291,6 +296,12 @@ func TestVolumes(t *testing.T) {
WaitForVolumeStatus(t, os, vol, volumeInUseStatus, volumeCreateTimeoutSeconds)
devicePath := os.GetDevicePath(diskId)
if !strings.HasPrefix(devicePath, "/dev/disk/by-id/") {
t.Fatalf("GetDevicePath returned and unexpected path for Cinder volume %s, returned %s", vol, devicePath)
}
t.Logf("Volume (%s) found at path: %s\n", vol, devicePath)
err = os.DetachDisk(os.localInstanceID, vol)
if err != nil {
t.Fatalf("Cannot DetachDisk Cinder volume %s: %v", vol, err)
......
......@@ -169,16 +169,25 @@ func (os *OpenStack) CreateVolume(name string, size int, vtype, availability str
// GetDevicePath returns the path of an attached block storage volume, specified by its id.
func (os *OpenStack) GetDevicePath(diskId string) string {
// Build a list of candidate device paths
candidateDeviceNodes := []string{
// KVM
fmt.Sprintf("virtio-%s", diskId[:20]),
// ESXi
fmt.Sprintf("wwn-0x%s", strings.Replace(diskId, "-", "", -1)),
}
files, _ := ioutil.ReadDir("/dev/disk/by-id/")
for _, f := range files {
if strings.Contains(f.Name(), "virtio-") {
devid_prefix := f.Name()[len("virtio-"):len(f.Name())]
if strings.Contains(diskId, devid_prefix) {
for _, c := range candidateDeviceNodes {
if c == f.Name() {
glog.V(4).Infof("Found disk attached as %q; full devicepath: %s\n", f.Name(), path.Join("/dev/disk/by-id/", f.Name()))
return path.Join("/dev/disk/by-id/", f.Name())
}
}
}
glog.Warningf("Failed to find device for the diskid: %q\n", diskId)
return ""
}
......@@ -208,8 +217,10 @@ func (os *OpenStack) DeleteVolume(volumeName string) error {
return err
}
// Get device path of attached volume to the compute running kubelet
// Get device path of attached volume to the compute running kubelet, as known by cinder
func (os *OpenStack) GetAttachmentDiskPath(instanceID string, diskName string) (string, error) {
// See issue #33128 - Cinder does not always tell you the right device path, as such
// we must only use this value as a last resort.
disk, err := os.getVolume(diskName)
if err != nil {
return "", err
......@@ -269,3 +280,8 @@ func (os *OpenStack) diskIsUsed(diskName string) (bool, error) {
}
return false, nil
}
// query if we should trust the cinder provide deviceName, See issue #33128
func (os *OpenStack) ShouldTrustDevicePath() bool {
return os.bsOpts.TrustDevicePath
}
......@@ -629,8 +629,10 @@ func (rs *Rackspace) DetachDisk(instanceID string, partialDiskId string) error {
return nil
}
// Get device path of attached volume to the compute running kubelet
// Get device path of attached volume to the compute running kubelet, as known by cinder
func (rs *Rackspace) GetAttachmentDiskPath(instanceID string, diskName string) (string, error) {
// See issue #33128 - Cinder does not always tell you the right device path, as such
// we must only use this value as a last resort.
disk, err := rs.getVolume(diskName)
if err != nil {
return "", err
......@@ -681,3 +683,8 @@ func (rs *Rackspace) DisksAreAttached(diskNames []string, instanceID string) (ma
}
return attached, returnedErr
}
// query if we should trust the cinder provide deviceName, See issue #33128
func (rs *Rackspace) ShouldTrustDevicePath() bool {
return true
}
......@@ -144,6 +144,7 @@ func (attacher *cinderDiskAttacher) VolumesAreAttached(specs []*volume.Spec, nod
}
func (attacher *cinderDiskAttacher) WaitForAttach(spec *volume.Spec, devicePath string, timeout time.Duration) (string, error) {
// NOTE: devicePath is is path as reported by Cinder, which may be incorrect and should not be used. See Issue #33128
volumeSource, _, err := getVolumeSource(spec)
if err != nil {
return "", err
......@@ -166,13 +167,17 @@ func (attacher *cinderDiskAttacher) WaitForAttach(spec *volume.Spec, devicePath
case <-ticker.C:
glog.V(5).Infof("Checking Cinder disk %q is attached.", volumeID)
probeAttachedVolume()
if !attacher.cinderProvider.ShouldTrustDevicePath() {
// Using the Cinder volume ID, find the real device path (See Issue #33128)
devicePath = attacher.cinderProvider.GetDevicePath(volumeID)
}
exists, err := volumeutil.PathExists(devicePath)
if exists && err == nil {
glog.Infof("Successfully found attached Cinder disk %q.", volumeID)
glog.Infof("Successfully found attached Cinder disk %q at %v.", volumeID, devicePath)
return devicePath, nil
} else {
//Log error, if any, and continue checking periodically
glog.Errorf("Error Stat Cinder disk (%q) is attached: %v", volumeID, err)
// Log an error, and continue checking periodically
glog.Errorf("Error: could not find attached Cinder disk %q: %v", volumeID, err)
}
case <-timer.C:
return "", fmt.Errorf("Could not find attached Cinder disk %q. Timeout waiting for mount paths to be created.", volumeID)
......
......@@ -393,6 +393,10 @@ func (testcase *testcase) GetAttachmentDiskPath(instanceID string, diskName stri
return expected.retPath, expected.ret
}
func (testcase *testcase) ShouldTrustDevicePath() bool {
return true
}
func (testcase *testcase) CreateVolume(name string, size int, vtype, availability string, tags *map[string]string) (volumeName string, err error) {
return "", errors.New("Not implemented")
}
......
......@@ -51,6 +51,7 @@ type CinderProvider interface {
GetAttachmentDiskPath(instanceID string, diskName string) (string, error)
DiskIsAttached(diskName, instanceID string) (bool, error)
DisksAreAttached(diskNames []string, instanceID string) (map[string]bool, error)
ShouldTrustDevicePath() bool
Instances() (cloudprovider.Instances, bool)
}
......
Markdown is supported
0% or
You are about to add 0 people to the discussion. Proceed with caution.
Finish editing this message first!
Please register or to comment