local-volume/

README

Local Persistent Storage User Guide

Overview

Local persistent volumes allows users to access local storage through the standard PVC interface in a simple and portable way. The PV contains node affinity information that the system uses to schedule pods to the correct nodes.

An external static provisioner is available to help simplify local storage management once the local volumes are configured. Note that the local storage provisioner is different from most provisioners and does not support dynamic provisioning yet. Instead, it requires that administrators preconfigure the local volumes on each node and if volumes are supposed to be

1. Filesystem volumeMode (default) PVs - mount them under discovery directories.
2. Block volumeMode PVs - create a symbolic link under discovery directory to the block device on the node.

The provisioner will manage the volumes under the discovery directories by creating and cleaning up PersistentVolumes for each volume.

Configuration Requirements

• The local-volume plugin expects paths to be stable, including across reboots and when disks are added or removed.
• The static provisioner only discovers either mount points (for Filesystem mode volumes) or symbolic links (for Block mode volumes). For directory-based local volumes, they must be bind-mounted into the discovery directories.

Version Compatibility

Recommended provisioner versions with Kubernetes versions

Provisioner version	K8s version	Reason
2.1.0	1.10	Beta API default, block
2.0.0	1.8, 1.9	Mount propagation
1.0.1	1.7

K8s Feature Status

Also see known issues and CHANGELOG.

1.10: Beta

• New PV.NodeAffinity field added.
• Important: Alpha PV NodeAffinity annotation is deprecated. Users must manually update their PVs to use the new NodeAffinity field or run a one-time update job.
• Alpha: Raw block support added.

1.9: Alpha

• New StorageClass volumeBindingMode parameter that will delay PVC binding until a pod is scheduled.

1.7: Alpha

• New local PersistentVolume source that allows specifying a directory or mount point with node affinity.
• Pod using the PVC that is bound to this PV will always get scheduled to that node.

Future features

• Local block devices as a volume source, with partitioning and fs formatting
• Dynamic provisioning for shared local persistent storage
• Local PV health monitoring, taints and tolerations
• Inline PV (use dedicated local disk as ephemeral storage)

User Guide

These instructions reflect the latest version of the codebase. For instructions on older versions, please see version links under Version Compatibility.

Step 1: Bringing up a cluster with local disks

Enabling the alpha feature gates

1.10+

If raw local block feature is needed,

$ export KUBE_FEATURE_GATES="BlockVolume=true"

Note: Kubernetes versions prior to 1.10 require several additional feature-gates be enabled on all Kubernetes components, because the persistent lcoal volumes and other features were in alpha.

Option 1: GCE

GCE clusters brought up with kube-up.sh will automatically format and mount the requested Local SSDs, so you can deploy the provisioner with the pre-generated deployment spec and skip to step 4, unless you want to customize the provisioner spec or storage classes.

$ NODE_LOCAL_SSDS_EXT=<n>,<scsi|nvme>,fs cluster/kube-up.sh
$ kubectl create -f provisioner/deployment/kubernetes/gce/class-local-ssds.yaml
$ kubectl create -f provisioner/deployment/kubernetes/gce/provisioner_generated_gce_ssd_volumes.yaml

Option 2: GKE

GKE clusters will automatically format and mount the requested Local SSDs. Please see GKE documentation for instructions for how to create a cluster with Local SSDs.

Then skip to step 4.

Note: The raw block feature is only supported on GKE Kubernetes alpha clusters.

Option 3: Baremetal environments

1. Partition and format the disks on each node according to your application's requirements.
2. Mount all the filesystems under one directory per StorageClass. The directories are specified in a configmap, see below.
3. Configure the Kubernetes API Server, controller-manager, scheduler, and all kubelets with KUBE_FEATURE_GATES as described above.
4. If not using the default Kubernetes scheduler policy, the following predicates must be enabled:
   • Pre-1.9: NoVolumeBindConflict
   • 1.9+: VolumeBindingChecker

Option 4: Local test cluster

1. Create /mnt/disks directory and mount several volumes into its subdirectories. The example below uses three ram disks to simulate real local volumes:

$ mkdir /mnt/disks
$ for vol in vol1 vol2 vol3; do
    mkdir /mnt/disks/$vol
    mount -t tmpfs $vol /mnt/disks/$vol
done

1. Run the local cluster.

$ ALLOW_PRIVILEGED=true LOG_LEVEL=5 FEATURE_GATES=$KUBE_FEATURE_GATES hack/local-up-cluster.sh

Step 2: Creating a StorageClass (1.9+)

To delay volume binding until pod scheduling and to handle multiple local PVs in a single pod, a StorageClass must to be created with volumeBindingMode set to WaitForFirstConsumer.

$ kubectl create -f provisioner/deployment/kubernetes/example/default_example_storageclass.yaml

Step 3: Creating local persistent volumes

Option 1: Using the local volume static provisioner

1. Generate Provisioner's ServiceAccount, Roles, DaemonSet, and ConfigMap spec, and customize it.

   This step uses helm templates to generate the specs. See the helm README for setup instructions. To generate the provisioner's specs using the default values, run:

   helm template ./helm/provisioner > ./provisioner/deployment/kubernetes/provisioner_generated.yaml
   

   You can also provide a custom values file instead:

   helm template ./helm/provisioner --values custom-values.yaml > ./provisioner/deployment/kubernetes/provisioner_generated.yaml
   

2. Deploy Provisioner

   Once a user is satisfied with the content of Provisioner's yaml file, kubectl can be used to create Provisioner's DaemonSet and ConfigMap.

   $ kubectl create -f ./provisioner/deployment/kubernetes/provisioner_generated.yaml
   

3. Check discovered local volumes

   Once launched, the external static provisioner will discover and create local-volume PVs.

   For example, if the directory /mnt/disks/ contained one directory /mnt/disks/vol1 then the following local-volume PV would be created by the static provisioner:

   $ kubectl get pv
   NAME                CAPACITY    ACCESSMODES   RECLAIMPOLICY   STATUS      CLAIM     STORAGECLASS    REASON    AGE
   local-pv-ce05be60   1024220Ki   RWO           Delete          Available             local-storage             26s
   
   $ kubectl describe pv local-pv-ce05be60 
   Name:		local-pv-ce05be60
   Labels:		<none>
   Annotations:	pv.kubernetes.io/provisioned-by=local-volume-provisioner-minikube-18f57fb2-a186-11e7-b543-080027d51893
   StorageClass:	local-fast
   Status:		Available
   Claim:		
   Reclaim Policy:	Delete
   Access Modes:	RWO
   Capacity:	1024220Ki
   NodeAffinity:
     Required Terms:
         Term 0:  kubernetes.io/hostname in [my-node]
   Message:	
   Source:
       Type:	LocalVolume (a persistent volume backed by local storage on a node)
       Path:	/mnt/disks/vol1
   Events:		<none>
   

   The PV described above can be claimed and bound to a PVC by referencing the local-fast storageClassName.

Option 2: Manually create local persistent volume

See Kubernetes documentation for an example PersistentVolume spec.

Step 4: Create local persistent volume claim

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: example-local-claim
spec:
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 5Gi
  storageClassName: local-storage

Please replace the following elements to reflect your configuration:

• "5Gi" with required size of storage volume
• "local-storage" with the name of storage class associated with the local PVs that should be used for satisfying this PVC

For "Block" volumeMode PVC, which tries to claim a "Block" PV, the following example can be used:

kind: PersistentVolumeClaim
apiVersion: v1
metadata:
  name: example-local-claim
spec:
  accessModes:
  - ReadWriteOnce
  resources:
    requests:
      storage: 5Gi
  volumeMode: Block
  storageClassName: local-storage

Note that the only additional field of interest here is volumeMode, which has been set to "Block".

E2E Tests

Running

go run hack/e2e.go -- -v --test --test_args="--ginkgo.focus=PersistentVolumes-local"

View CI Results

GCE

GKE

GCE Slow

GKE Slow

GCE Serial

GKE Serial

GCE Alpha

GKE Alpha

Best Practices

• For IO isolation, a whole disk per volume is recommended
• For capacity isolation, separate partitions per volume is recommended
• Avoid recreating nodes with the same node name while there are still old PVs with that node's affinity specified. Otherwise, the system could think that the new node contains the old PVs.
• For volumes with a filesystem, it's recommended to utilize their UUID (e.g. the output from ls -l /dev/disk/by-uuid) both in fstab entries and in the directory name of that mount point. This practice ensures that the wrong local volume is not mistakenly mounted, even if its device path changes (e.g. if /dev/sda1 becomes /dev/sdb1 when a new disk is added). Additionally, this practice will ensure that if another node with the same name is created, that any volumes on that node are unique and not mistaken for a volume on another node with the same name.
• For raw block volumes without a filesystem, use a unique ID as the symlink name. Depending on your environment, the volume's ID in /dev/disk/by-id/ may contain a unique hardware serial number. Otherwise, a unique ID should be generated. The uniqueness of the symlink name will ensure that if another node with the same name is created, that any volumes on that node are unique and not mistaken for a volume on another node with the same name.

Deleting/removing the underlying volume

When you want to decommission the local volume, here is a possible workflow.

1. Stop the pods that are using the volume
2. Remove the local volume from the node (ie unmounting, pulling out the disk, etc)
3. Delete the PVC
4. The provisioner will try to cleanup the volume, but will fail since the volume no longer exists
5. Manually delete the PV object