There is a race in dynamic provisioning:
1. User creates a claim to dynamically provision a volume.
2. Kubernetes start provisioning the volume.
3. User deletes the claim before 2. is finished.
4. Kubernetes finished provisioning the volume.
The volume is in Pending state. This volume should be deleted instead of
moving to Available state.
Volume names have now format <cluster-name>-dynamic-<pv-name>.
pv-name is guaranteed to be unique in Kubernetes cluster, adding
<cluster-name> ensures we don't conflict with any running cluster
in the cloud project (kube-controller-manager --cluster-name=XXX).
'kubernetes' is the default cluster name.
Recycle controller tries to recycle or delete a PV several times.
It stores count of failed attempts and timestamp of the last attempt in
annotations of the PV.
By default, the controller tries to recycle/delete a PV 3 times in
10 minutes interval. These values are configurable by
kube-controller-manager --pv-recycler-maximum-retry=X --pvclaimbinder-sync-period=Y
arguments.
Fixes#19860 (it may be easier to look at the issue to see exact sequence
to reproduce the bug and understand the fix).
When PersistentVolumeProvisionerController.reconcileClaim() is called with the
same claim in short succession (e.g. the claim is created by an user and
at the same time periodic check of all claims is scheduled), the second
reconcileClaim() call gets an old copy of the claim as its parameter.
The method should always reload the claim to get a fresh copy with all
annotations, possibly added by previous reconcileClaim() call.
The same applies to PersistentVolumeClaimBinder.syncClaim().
Also update all the test to store claims in "fake" API server before calling
syncClaim and reconcileClaim.
We should always load the newest version of the volume from APIserver before
processing it.
When PersistentVolumeProvisionerController.reconcileVolume() is called with the
same volume in short succession (e.g. the volume is created by a provisioner
and at the same time periodic check of all volumes is scheduled), the second
reconcileVolume() call gets an old copy of the volume as its parameter and
it does not see annotations updated by the previous call.
This may result in one volume being provisioned several times, creating orphan
volumes in the cloud.
The same error is in PersistentVolumeClaimBinder.syncVolume().
We add claim.Name and claim.Namespace as tags to AWS EBS / GCE PD / OpenStack
Cinder volumes created by Kubernetes. To easily match Kubernetes volumes and
cloud volumes, let's add also PV.Name.
- Add volume.MetricsProvider function to Volume interface.
- Add volume.MetricsDu for providing metrics via executing "du".
- Add volulme.MetricsNil for unsupported Volumes.
