Merge pull request #62612 from andyzhangx/azure-devicepath-fix

Automatic merge from submit-queue (batch tested with PRs 62676, 62612). If you want to cherry-pick this change to another branch, please follow the instructions <a href="https://github.com/kubernetes/community/blob/master/contributors/devel/cherry-picks.md">here</a>. fix WaitForAttach failure issue for azure disk **What this PR does / why we need it**: From v1.10, `devicePath` will be updated due to following code change: https://github.com/kubernetes/kubernetes/blob/568afb4ecca99bc3b54fddc20927b3713369d357/pkg/volume/util/operationexecutor/operation_generator.go#L517-L518 So in v1.10.0, MountVolume.WaitForAttach will fail in the azure disk remount, error logs would be like following: ``` MountVolume.WaitForAttach failed for volume "pvc-f1562ecb-3e5f-11e8-ab6b-000d3af9f967" : azureDisk - Wait for attach expect device path as a lun number, instead got: /dev/disk/azure/scsi1/lun1 (strconv.Atoi: parsing "/dev/disk/azure/scsi1/lun1": invalid syntax) Warning FailedMount 1m (x10 over 21m) kubelet, k8s-agentpool-66825246-0 Unable to mount volumes for pod ``` This PR does not use `devicePath` anymore since it could be changed, instead, it use `diskController.GetDiskLun(diskName, volumeSource.DataDiskURI, nodeName)` to get disk LUN, this ARM api call would cost about 0.12s The GCE disk won't have this issue since `devicePath` is not used in [WaitForAttach func](https://github.com/kubernetes/kubernetes/blob/master/pkg/volume/gce_pd/attacher.go#L133), while aws disk is also using `devicePath` in [WaitForAttach func](https://github.com/kubernetes/kubernetes/blob/master/pkg/volume/aws_ebs/attacher.go#L145), I think there is potentical issue for aws_ebs **Which issue(s) this PR fixes** *(optional, in `fixes #<issue number>(, fixes #<issue_number>, ...)` format, will close the issue(s) when PR gets merged)*: Fixes #62540 **Special notes for your reviewer**: should cherry-pick to v1.10 **Release note**: ``` fix WaitForAttach failure issue for azure disk ``` /assign @feiskyer /sig azure FYI @khenidak

Merge pull request #62612 from andyzhangx/azure-devicepath-fix
bf3cda66 · Kubernetes Submit Queue · GitHub · a73c80a7 · 3a0fb103 · bf3cda66
Unverified Commit bf3cda66 authored Apr 16, 2018 by Kubernetes Submit Queue Committed by GitHub Apr 16, 2018
Hide whitespace changes
Inline Side-by-side

Showing with 15 additions and 6 deletions

attacher.go pkg/volume/azure_dd/attacher.go +15 -6

No files found.
--- a/pkg/volume/azure_dd/attacher.go
+++ b/pkg/volume/azure_dd/attacher.go
@@ -150,27 +150,36 @@ func (a *azureDiskAttacher) VolumesAreAttached(specs []*volume.Spec, nodeName ty
 func (a *azureDiskAttacher) WaitForAttach(spec *volume.Spec, devicePath string, _ *v1.Pod, timeout time.Duration) (string, error) {
 	var err error
-	lun, err := strconv.Atoi(devicePath)
+	volumeSource, err := getVolumeSource(spec)
 	if err != nil {
-		return "", fmt.Errorf("azureDisk - Wait for attach expect device path as a lun number, instead got: %s (%v)", devicePath, err)
+		return "", err
 	}
-	volumeSource, err := getVolumeSource(spec)
+	diskController, err := getDiskController(a.plugin.host)
 	if err != nil {
 		return "", err
 	}
+	nodeName := types.NodeName(a.plugin.host.GetHostName())
+	diskName := volumeSource.DiskName
+	glog.V(5).Infof("azureDisk - WaitForAttach: begin to GetDiskLun by diskName(%s), DataDiskURI(%s), nodeName(%s), devicePath(%s)",
+		diskName, volumeSource.DataDiskURI, nodeName, devicePath)
+	lun, err := diskController.GetDiskLun(diskName, volumeSource.DataDiskURI, nodeName)
+	if err != nil {
+		return "", err
+	}
+	glog.V(5).Infof("azureDisk - WaitForAttach: GetDiskLun succeeded, got lun(%v)", lun)
 	exec := a.plugin.host.GetExec(a.plugin.GetPluginName())
 	io := &osIOHandler{}
 	scsiHostRescan(io, exec)
-	diskName := volumeSource.DiskName
-	nodeName := a.plugin.host.GetHostName()
 	newDevicePath := ""
 	err = wait.Poll(1*time.Second, timeout, func() (bool, error) {
-		if newDevicePath, err = findDiskByLun(lun, io, exec); err != nil {
+		if newDevicePath, err = findDiskByLun(int(lun), io, exec); err != nil {
 			return false, fmt.Errorf("azureDisk - WaitForAttach ticker failed node (%s) disk (%s) lun(%v) err(%s)", nodeName, diskName, lun, err)
 		}