github/kubernetes
1
0
Fork 0
mirror of https://github.com/k3s-io/kubernetes.git synced 2025-08-13 13:55:41 +00:00
Code Issues Projects Releases Wiki Activity

kubelet: Fix the volume manager did't check the device mount state in the actual state of the world before marking the volume as detached. It may cause a pod to be stuck in the Terminating state due to the above issue when it was deleted.

Browse Source
This commit is contained in:
carlory 2024-10-21 15:01:45 +08:00
parent 9b2866fccb
commit 780ebfa496
2 changed files with 22 additions and 0 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

17
pkg/kubelet/volumemanager/cache/actual_state_of_world.go vendored
View File

@ -168,6 +168,11 @@ type ActualStateOfWorld interface {
// or have a mount/unmount operation pending. // or have a mount/unmount operation pending.
GetAttachedVolumes() []AttachedVolume GetAttachedVolumes() []AttachedVolume
// GetAttachedVolume returns the volume that is known to be attached to the node
// with the given volume name. If the volume is not found, the second return value
// is false.
GetAttachedVolume(volumeName v1.UniqueVolumeName) (AttachedVolume, bool)
// SyncReconstructedVolume check the volume.outerVolumeSpecName in asw and // SyncReconstructedVolume check the volume.outerVolumeSpecName in asw and
// the one populated from dsw, if they do not match, update this field from the value from dsw. // the one populated from dsw, if they do not match, update this field from the value from dsw.
SyncReconstructedVolume(volumeName v1.UniqueVolumeName, podName volumetypes.UniquePodName, outerVolumeSpecName string) SyncReconstructedVolume(volumeName v1.UniqueVolumeName, podName volumetypes.UniquePodName, outerVolumeSpecName string)
@ -1104,6 +1109,18 @@ func (asw *actualStateOfWorld) GetAttachedVolumes() []AttachedVolume {
return allAttachedVolumes return allAttachedVolumes
} }
func (asw *actualStateOfWorld) GetAttachedVolume(volumeName v1.UniqueVolumeName) (AttachedVolume, bool) {
asw.RLock()
defer asw.RUnlock()
volumeObj, ok := asw.attachedVolumes[volumeName]
if !ok {
return AttachedVolume{}, false
}
return asw.newAttachedVolume(&volumeObj), true
}
func (asw *actualStateOfWorld) GetUnmountedVolumes() []AttachedVolume { func (asw *actualStateOfWorld) GetUnmountedVolumes() []AttachedVolume {
asw.RLock() asw.RLock()
defer asw.RUnlock() defer asw.RUnlock()

5
pkg/kubelet/volumemanager/reconciler/reconciler_common.go
View File

@ -270,6 +270,11 @@ func (rc *reconciler) unmountDetachDevices() {
// Check IsOperationPending to avoid marking a volume as detached if it's in the process of mounting. // Check IsOperationPending to avoid marking a volume as detached if it's in the process of mounting.
if !rc.desiredStateOfWorld.VolumeExists(attachedVolume.VolumeName, attachedVolume.SELinuxMountContext) && if !rc.desiredStateOfWorld.VolumeExists(attachedVolume.VolumeName, attachedVolume.SELinuxMountContext) &&
!rc.operationExecutor.IsOperationPending(attachedVolume.VolumeName, nestedpendingoperations.EmptyUniquePodName, nestedpendingoperations.EmptyNodeName) { !rc.operationExecutor.IsOperationPending(attachedVolume.VolumeName, nestedpendingoperations.EmptyUniquePodName, nestedpendingoperations.EmptyNodeName) {
// Re-read the actual state of the world, maybe the volume got mounted in the meantime.
// This is safe, because there is no pending operation (checked above) and no new operation
// could start in the meantime. The only goroutine that adds new operations is this reconciler.
attachedVolume, _ = rc.actualStateOfWorld.GetAttachedVolume(attachedVolume.VolumeName)
if attachedVolume.DeviceMayBeMounted() { if attachedVolume.DeviceMayBeMounted() {
// Volume is globally mounted to device, unmount it // Volume is globally mounted to device, unmount it
klog.V(5).InfoS(attachedVolume.GenerateMsgDetailed("Starting operationExecutor.UnmountDevice", "")) klog.V(5).InfoS(attachedVolume.GenerateMsgDetailed("Starting operationExecutor.UnmountDevice", ""))