Longhorn

Longhorn is a distributed block storage system for Kubernetes. Longhorn is lightweight, reliable, and easy-to-use. You can deploy Longhorn on an existing Kubernetes cluster with one simple command. Once Longhorn is deployed, it adds persistent volume support to the Kubernetes cluster.

Longhorn implements distributed block storage using containers and microservices. Longhorn creates a dedicated storage controller for each block device volume and sychronously replicates the volume across multiple replicas stored on multiple nodes. The storage controller and replicas are themselves orchestrated using Kubernetes. Longhorn supports snapshots, backups, and even allows you to schedule recurring snapshots and backups!

You can read more details of Longhorn and its design here.

Longhorn is a work in progress. We appreciate your comments as we continue to work on it!

Source Code

Longhorn is 100% open source software. Project source code is spread across a number of repos:

1. Longhorn Engine -- Core controller/replica logic https://github.com/rancher/longhorn-engine
2. Longhorn Manager -- Longhorn orchestration, includes Flexvolume driver for Kubernetes https://github.com/rancher/longhorn-manager
3. Longhorn UI -- Dashboard https://github.com/rancher/longhorn-ui

Demo

Requirements

Minimal Requirements

1. Docker v1.13+
2. Kubernetes v1.8+
3. Make sure open-iscsi has been installed in all nodes of the Kubernetes cluster. For GKE, recommended Ubuntu as guest OS image since it contains open-iscsi already.

Kubernetes Driver Requirements

Longhorn can be used in Kubernetes to provide persistent storage through either Longhorn Container Storage Interface (CSI) driver or Longhorn Flexvolume driver. Longhorn will automatically deploy one of the drivers, depends on user's Kubernetes cluster's setup. User can also specify the driver in the deployment yaml file. CSI is preferred.

Requirement for the CSI driver

1. Kubernetes v1.10+ 1.1 CSI is in beta release for this version of Kubernetes, and enabled by default.
2. Mount Propagation feature gate enabled. 2.1 It's enabled by default in Kubernetes v1.10. But some early versions of RKE may not enable it.
3. If above conditions cannot be met, Longhorn will falls back to use Flexvolume driver.

Requirement for the Flexvolume driver

1. Kubernetes v1.8+
2. Make sure curl, findmnt, grep, awk and blkid has been installed in the every node of the Kubernetes cluster.
3. User need to know the volume plugin directory in order to setup the driver correctly.
   1. Rancher RKE: /var/lib/kubelet/volumeplugins
   2. Google GKE: /home/kubernetes/flexvolume
   3. For other distro, please find the correct directory by running ps aux|grep kubelet on the host and check the --volume-plugin-dir parameter. If there is none, it would be the default value /usr/libexec/kubernetes/kubelet-plugins/volume/exec/ .

Deployment

Create the deployment of Longhorn in your Kubernetes cluster is easy.

If you're using Rancher RKE, or other distro with Kubernetes v1.10+ and Mount Propagation enabled, you can just do:

kubectl create -f https://raw.githubusercontent.com/rancher/longhorn/rc/deploy/longhorn.yaml

If you're using Flexvolume driver with other Kubernetes Distro, replace the value of $FLEXVOLUME_DIR in the following command with your own Flexvolume Directory as specified above.

FLEXVOLUME_DIR="/home/kubernetes/flexvolume/"
curl -s https://raw.githubusercontent.com/rancher/longhorn/rc/deploy/longhorn.yaml|sed "s#^\( *\)value: \"/var/lib/kubelet/volumeplugins\"#\1value: \"${FLEXVOLUME_DIR}\"#g" > longhorn.yaml
kubectl create -f longhorn.yaml

For Google Kubernetes Engine (GKE) users, see here before proceed.

Longhorn Manager and Longhorn Driver will be deployed as daemonsets in a separate namespace called longhorn-system, as you can see in the yaml file.

When you see those pods has started correctly as follows, you've deployed the Longhorn successfully.

Deployed with CSI driver:

# kubectl -n longhorn-system get pod
NAME                                        READY     STATUS    RESTARTS   AGE
csi-attacher-0                              1/1       Running   0          6h
csi-provisioner-0                           1/1       Running   0          6h
engine-image-ei-57b85e25-8v65d              1/1       Running   0          7d
engine-image-ei-57b85e25-gjjs6              1/1       Running   0          7d
engine-image-ei-57b85e25-t2787              1/1       Running   0          7d
longhorn-csi-plugin-4cpk2                   2/2       Running   0          6h
longhorn-csi-plugin-ll6mq                   2/2       Running   0          6h
longhorn-csi-plugin-smlsh                   2/2       Running   0          6h
longhorn-driver-deployer-7b5bdcccc8-fbncl   1/1       Running   0          6h
longhorn-manager-7x8x8                      1/1       Running   0          6h
longhorn-manager-8kqf4                      1/1       Running   0          6h
longhorn-manager-kln4h                      1/1       Running   0          6h
longhorn-ui-f849dcd85-cgkgg                 1/1       Running   0          5d

Or with Flexvolume driver

# kubectl -n longhorn-system get pod
NAME                                        READY     STATUS    RESTARTS   AGE
engine-image-ei-57b85e25-8v65d              1/1       Running   0          7d
engine-image-ei-57b85e25-gjjs6              1/1       Running   0          7d
engine-image-ei-57b85e25-t2787              1/1       Running   0          7d
longhorn-driver-deployer-5469b87b9c-b9gm7   1/1       Running   0          2h
longhorn-flexvolume-driver-lth5g            1/1       Running   0          2h
longhorn-flexvolume-driver-tpqf7            1/1       Running   0          2h
longhorn-flexvolume-driver-v9mrj            1/1       Running   0          2h
longhorn-manager-7x8x8                      1/1       Running   0          9h
longhorn-manager-8kqf4                      1/1       Running   0          9h
longhorn-manager-kln4h                      1/1       Running   0          9h
longhorn-ui-f849dcd85-cgkgg                 1/1       Running   0          5d

Access the UI

Use kubectl -n longhorn-system get svc to get the external service IP for UI:

NAME                TYPE           CLUSTER-IP      EXTERNAL-IP      PORT(S)        AGE
longhorn-backend    ClusterIP      10.20.248.250   <none>           9500/TCP       58m
longhorn-frontend   LoadBalancer   10.20.245.110   100.200.200.123   80:30697/TCP   58m

If the Kubernetes Cluster supports creating LoadBalancer, user can then use EXTERNAL-IP(100.200.200.123 in the case above) of longhorn-frontend to access the Longhorn UI. Otherwise the user can use <node_ip>:<port> (port is 30697in the case above) to access the UI.

Longhorn UI would connect to the Longhorn Manager API, provides the overview of the system, the volume operations, and the snapshot/backup operations. It's highly recommended for the user to check out Longhorn UI.

Notice the current UI is unauthenticated.

Use the Longhorn with Kubernetes

Longhorn provides persistent volume directly to Kubernetes through one of the Longhorn drivers. No matter which driver you're using, you can use Kubernetes StorageClass to provision your persistent volumes.

Use following command to create a default Longhorn StorageClass named longhorn.

kubectl create -f https://raw.githubusercontent.com/rancher/longhorn/rc/examples/storageclass.yaml

Then user can create a PVC directly. For example:

apiVersion: v1
kind: PersistentVolumeClaim
metadata:
  name: longhorn-volv-pvc
spec:
  accessModes:
    - ReadWriteOnce
  storageClassName: longhorn
  resources:
    requests:
      storage: 2Gi

Then use it in the pod:

apiVersion: v1
kind: Pod
metadata:
  name: volume-test
  namespace: default
spec:
  containers:
  - name: volume-test
    image: nginx:stable-alpine
    imagePullPolicy: IfNotPresent
    volumeMounts:
    - name: volv
      mountPath: /data
    ports:
    - containerPort: 80
  volumes:
  - name: volv
    persistentVolumeClaim:
      claimName: longhorn-volv-pvc

More examples are available at ./examples/

Feature Usage

Snapshot

A snapshot in Longhorn represents a volume state at a given time, stored in the same location of volume data on physical disk of the host. Snapshot creation is instant in Longhorn.

User can revert to any previous taken snapshot using the UI. Since Longhorn is a distributed block storage, please make sure the Longhorn volume is umounted from the host when revert to any previous snapshot, otherwise it will confuse the node filesystem and cause corruption.

Backup

A backup in Longhorn represents a volume state at a given time, stored in the BackupStore which is outside of the Longhorn System. Backup creation will involving copying the data through the network, so it will take time.

A corresponding snapshot is needed for creating a backup. And user can choose to backup any snapshot previous created.

A BackupStore is a NFS server or S3 compatible server.

A BackupTarget represents a BackupStore in the Longhorn System. The BackupTarget can be set at Settings/General/BackupTarget

If user is using a S3 compatible server as the BackupTarget, the BackupTargetSecret is needed for authentication informations. User need to manually create it as a Kubernetes Secret in the longhorn-system namespace. See below for details.

Setup a testing backupstore

We provides two testing purpose backupstore based on NFS server and Minio S3 server for testing, in ./deploy/backupstores.

Use following command to setup a Minio S3 server for BackupStore after longhorn-system was created.

kubectl -f https://raw.githubusercontent.com/rancher/longhorn/rc/deploy/backupstores/minio-backupstore.yaml

Now set Settings/General/BackupTarget to

http://minio-service.default:9000

And Setttings/General/BackupTargetSecret to

minio-secret

Click the Backup tab in the UI, it should report an empty list without error out.

Recurring Snapshot and Backup

Longhorn supports recurring snapshot and backup for volumes. User only need to set when he/she wish to take the snapshot and/or backup, and how many snapshots/backups needs to be retains, then Longhorn will automatically create snapshot/backup for the user at that time, as long as the volume is attached to a node.

User can find the setting for the recurring snapshot and backup in the Volume Detail page.

Multiple disks support

Longhorn supports to use more than one disk on the nodes to store the volume data.

To add a new disk for a node, heading to Node tab, select one of the node, and click the edit disk icon.

By default, /var/lib/rancher/longhorn on the host will be used for storing the volume data.

To add any additional disks, user needs to:

1. Mount the disk on the host to a certain directory.
2. Add the path of the mounted disk into the disk list of the node.

Longhorn will detect the storage information (e.g. maximum space, available space) about the disk automatically, and start scheduling to it if it's possible to accomodate the volume in there. A path mounted by the existing disk won't be allowed.

User can reserve a certain amount of space of the disk to stop Longhorn from using it. It can be set in the Space Reserved field for the disk. It's useful for the non-dedicated storage disk on the node.

Nodes and disks can be excluded from future scheduling. Notice any scheduled storage space won't be released automatically if the scheduling was disabled for the node.

There are two global settings affect the scheduling of the volume as well.

StorageOverProvisioningPercentage defines the upper bound of ScheduledStorage / (MaximumStorage - ReservedStorage) . The default value is 500 (%). That means we can schedule a total of 750 GiB Longhorn volumes on a 200 GiB disk with 50G reserved for the root file system. Because normally people won't use that large amount of data in the volume, and we store the volumes as sparse files.

StorageMinimalAvailablePercentage defines when a disk cannot be scheduled with more volumes. The default value is 10 (%). The bigger value between MaximumStorage * StorageMinimalAvailablePercentage / 100 and MaximumStorage - ReservedStorage will be used to determine if a disk is running low and cannot be scheduled with more volumes.

Notice currently there is no guarantee that the space volumes used won't exceed the StorageMinimalAvailablePercentage, because:

1. Longhorn volume can be bigger than specified size, due to the snapshot contains the old state of the volume
2. And Longhorn is doing over-provisioning by default.

Uninstall Longhorn

Longhorn CRD has finalizers in them, so user should delete the volumes and related resource first, give manager a chance to clean up after them.

1. Clean up volume and related resources

kubectl -n longhorn-system delete volumes.longhorn.rancher.io --all

Check the result using:

kubectl -n longhorn-system get volumes.longhorn.rancher.io
kubectl -n longhorn-system get engines.longhorn.rancher.io
kubectl -n longhorn-system get replicas.longhorn.rancher.io

Make sure all reports No resources found. before continuing.

2. Clean up engine images and nodes

kubectl -n longhorn-system delete engineimages.longhorn.rancher.io --all
kubectl -n longhorn-system delete nodes.longhorn.rancher.io --all

Check the result using:

kubectl -n longhorn-system get engineimages.longhorn.rancher.io
kubectl -n longhorn-system get nodes.longhorn.rancher.io

Make sure all reports No resources found. before continuing.

3. Uninstall Longhorn System

kubectl delete -f https://raw.githubusercontent.com/rancher/longhorn/rc/deploy/longhorn.yaml

Notes

Google Kubernetes Engine

The configuration yaml will be slight different for Google Kubernetes Engine (GKE):

1. GKE requires user to manually claim himself as cluster admin to enable RBAC. User need to execute following command before create the Longhorn system using yaml files.

kubectl create clusterrolebinding cluster-admin-binding --clusterrole=cluster-admin --user=<name@example.com>

In which name@example.com is the user's account name in GCE, and it's case sensitive. See here for details.

2. The default Flexvolume plugin directory is different with GKE 1.8+, which is at /home/kubernetes/flexvolume. User need to use following command instead:

FLEXVOLUME_DIR="/home/kubernetes/flexvolume/"
curl -s https://raw.githubusercontent.com/rancher/longhorn/rc/deploy/longhorn.yaml|sed "s#^\( *\)value: \"/var/lib/kubelet/volumeplugins\"#\1value: \"${FLEXVOLUME_DIR}\"#g" > longhorn.yaml
kubectl create -f longhorn.yaml

See Troubleshooting for details.

Troubleshooting

Volume can be attached/detached from UI, but Kubernetes Pod/StatefulSet etc cannot use it

Check if volume plugin directory has been set correctly.

By default, Kubernetes use /usr/libexec/kubernetes/kubelet-plugins/volume/exec/ as the directory for volume plugin drivers, as stated in the official document.

But some vendors may choose to change the directory due to various reasons. For example, GKE uses /home/kubernetes/flexvolume, and RKE uses /var/lib/kubelet/volumeplugins.

User can find the correct directory by running ps aux|grep kubelet on the host and check the --volume-plugin-dirparameter. If there is none, the default /usr/libexec/kubernetes/kubelet-plugins/volume/exec/ will be used.

License

Copyright (c) 2014-2018 Rancher Labs, Inc.

Licensed under the Apache License, Version 2.0 (the "License"); you may not use this file except in compliance with the License. You may obtain a copy of the License at

http://www.apache.org/licenses/LICENSE-2.0

Unless required by applicable law or agreed to in writing, software distributed under the License is distributed on an "AS IS" BASIS, WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied. See the License for the specific language governing permissions and limitations under the License.