Restic Integration

Velero supports backing up and restoring Kubernetes volumes using a free open-source backup tool called restic. This support is considered beta quality. Please see the list of limitations to understand if it fits your use case.

Velero allows you to take snapshots of persistent volumes as part of your backups if you’re using one of the supported cloud providers’ block storage offerings (Amazon EBS Volumes, Azure Managed Disks, Google Persistent Disks). It also provides a plugin model that enables anyone to implement additional object and block storage backends, outside the main Velero repository.

Velero’s Restic integration was added to give you an out-of-the-box solution for backing up and restoring almost any type of Kubernetes volume. This integration is an addition to Velero’s capabilities, not a replacement for existing functionality. If you’re running on AWS, and taking EBS snapshots as part of your regular Velero backups, there’s no need to switch to using Restic. However, if you need a volume snapshot plugin for your storage platform, or if you’re using EFS, AzureFile, NFS, emptyDir, local, or any other volume type that doesn’t have a native snapshot concept, Restic might be for you.

Restic is not tied to a specific storage platform, which means that this integration also paves the way for future work to enable cross-volume-type data migrations.

NOTE: hostPath volumes are not supported, but the local volume type is supported.

Setup Restic

Prerequisites

• Understand how Velero performs backups with the Restic integration.
• Download the latest Velero release.
• Kubernetes v1.16.0 and later. Velero’s Restic integration requires the Kubernetes MountPropagation feature.

Install Restic

To install Restic, use the --use-restic flag in the velero install command. See the install overview for more details on other flags for the install command.

velero install --use-restic

When using Restic on a storage provider that doesn’t have Velero support for snapshots, the --use-volume-snapshots=false flag prevents an unused VolumeSnapshotLocation from being created on installation.

Velero handles the creation of the restic repo prefix for Amazon, Azure, and GCP plugins, if you are using a different provider plugin, then you will need to make sure the resticRepoPrefix is set in the BackupStorageLocation config. The value for resticRepoPrefix should be the cloud storage URL where all namespace restic repos will be created. Velero creates one restic repo per namespace. For example, if backing up 2 namespaces, namespace1 and namespace2, using restic on AWS, the resticRepoPrefix would be something like s3:s3-us-west-2.amazonaws.com/bucket/restic and the full restic repo path for namespace1 would be s3:s3-us-west-2.amazonaws.com/bucket/restic/ns1 and for namespace2 would be s3:s3-us-west-2.amazonaws.com/bucket/restic/ns2.

There may be additional installation steps depending on the cloud provider plugin you are using. You should refer to the plugin specific documentation for the must up to date information.

Configure Restic DaemonSet spec

After installation, some PaaS/CaaS platforms based on Kubernetes also require modifications the Restic DaemonSet spec. The steps in this section are only needed if you are installing on RancherOS, OpenShift, VMware Tanzu Kubernetes Grid Integrated Edition (formerly VMware Enterprise PKS), or Microsoft Azure.

RancherOS

Update the host path for volumes in the Restic DaemonSet in the Velero namespace from /var/lib/kubelet/pods to /opt/rke/var/lib/kubelet/pods.

hostPath:
  path: /var/lib/kubelet/pods

to

hostPath:
  path: /opt/rke/var/lib/kubelet/pods

OpenShift

To mount the correct hostpath to pods volumes, run the Restic pod in privileged mode.

1. Add the velero ServiceAccount to the privileged SCC:

   $ oc adm policy add-scc-to-user privileged -z velero -n velero
   

2. For OpenShift version >= 4.1, modify the DaemonSet yaml to request a privileged mode:

   @@ -67,3 +67,5 @@ spec:
                 value: /credentials/cloud
               - name: VELERO_SCRATCH_DIR
                 value: /scratch
   +          securityContext:
   +            privileged: true
   

   or

   oc patch ds/restic \
     --namespace velero \
     --type json \
     -p '[{"op":"add","path":"/spec/template/spec/containers/0/securityContext","value": { "privileged": true}}]'
   

3. For OpenShift version < 4.1, modify the DaemonSet yaml to request a privileged mode and mount the correct hostpath to pods volumes.

   @@ -35,7 +35,7 @@ spec:
               secretName: cloud-credentials
           - name: host-pods
             hostPath:
   -            path: /var/lib/kubelet/pods
   +            path: /var/lib/origin/openshift.local.volumes/pods
           - name: scratch
             emptyDir: {}
         containers:
   @@ -67,3 +67,5 @@ spec:
                 value: /credentials/cloud
               - name: VELERO_SCRATCH_DIR
                 value: /scratch
   +          securityContext:
   +            privileged: true
   

   or

   oc patch ds/restic \
     --namespace velero \
     --type json \
     -p '[{"op":"add","path":"/spec/template/spec/containers/0/securityContext","value": { "privileged": true}}]'
   
   oc patch ds/restic \
     --namespace velero \
     --type json \
     -p '[{"op":"replace","path":"/spec/template/spec/volumes/0/hostPath","value": { "path": "/var/lib/origin/openshift.local.volumes/pods"}}]'
   

If Restic is not running in a privileged mode, it will not be able to access pods volumes within the mounted hostpath directory because of the default enforced SELinux mode configured in the host system level. You can create a custom SCC to relax the security in your cluster so that Restic pods are allowed to use the hostPath volume plug-in without granting them access to the privileged SCC.

By default a userland openshift namespace will not schedule pods on all nodes in the cluster.

To schedule on all nodes the namespace needs an annotation:

oc annotate namespace <velero namespace> openshift.io/node-selector=""

This should be done before velero installation.

Or the ds needs to be deleted and recreated:

oc get ds restic -o yaml -n <velero namespace> > ds.yaml
oc annotate namespace <velero namespace> openshift.io/node-selector=""
oc create -n <velero namespace> -f ds.yaml

VMware Tanzu Kubernetes Grid Integrated Edition (formerly VMware Enterprise PKS)

You need to enable the Allow Privileged option in your plan configuration so that Restic is able to mount the hostpath.

The hostPath should be changed from /var/lib/kubelet/pods to /var/vcap/data/kubelet/pods

hostPath:
  path: /var/vcap/data/kubelet/pods

Microsoft Azure

If you are using Azure Files, you need to add nouser_xattr to your storage class’s mountOptions. See this restic issue for more details.

You can use the following command to patch the storage class:

kubectl patch storageclass/<YOUR_AZURE_FILE_STORAGE_CLASS_NAME> \
  --type json \
  --patch '[{"op":"add","path":"/mountOptions/-","value":"nouser_xattr"}]'

To back up

Velero supports two approaches of discovering pod volumes that need to be backed up using Restic:

• Opt-in approach: Where every pod containing a volume to be backed up using Restic must be annotated with the volume’s name.
• Opt-out approach: Where all pod volumes are backed up using Restic, with the ability to opt-out any volumes that should not be backed up.

The following sections provide more details on the two approaches.

Using the opt-out approach

In this approach, Velero will back up all pod volumes using Restic with the exception of:

• Volumes mounting the default service account token, Kubernetes Secrets, and ConfigMaps
• Hostpath volumes

It is possible to exclude volumes from being backed up using the backup.velero.io/backup-volumes-excludes annotation on the pod.

Instructions to back up using this approach are as follows:

1. Run the following command on each pod that contains volumes that should not be backed up using Restic

   kubectl -n YOUR_POD_NAMESPACE annotate pod/YOUR_POD_NAME backup.velero.io/backup-volumes-excludes=YOUR_VOLUME_NAME_1,YOUR_VOLUME_NAME_2,...
   

   where the volume names are the names of the volumes in the pod spec.

   For example, in the following pod:

   apiVersion: v1
   kind: Pod
   metadata:
     name: app1
     namespace: sample
   spec:
     containers:
     - image: k8s.gcr.io/test-webserver
       name: test-webserver
       volumeMounts:
       - name: pvc1-vm
         mountPath: /volume-1
       - name: pvc2-vm
         mountPath: /volume-2
     volumes:
     - name: pvc1-vm
       persistentVolumeClaim:
         claimName: pvc1
     - name: pvc2-vm
         claimName: pvc2
   

   to exclude Restic backup of volume pvc1-vm, you would run:

   kubectl -n sample annotate pod/app1 backup.velero.io/backup-volumes-excludes=pvc1-vm
   

2. Take a Velero backup:

   velero backup create BACKUP_NAME --default-volumes-to-restic OTHER_OPTIONS
   

   The above steps uses the opt-out approach on a per backup basis.

   Alternatively, this behavior may be enabled on all velero backups running the velero install command with the --default-volumes-to-restic flag. Refer install overview for details.

3. When the backup completes, view information about the backups:

   velero backup describe YOUR_BACKUP_NAME
   

   kubectl -n velero get podvolumebackups -l velero.io/backup-name=YOUR_BACKUP_NAME -o yaml
   

Using opt-in pod volume backup

Velero, by default, uses this approach to discover pod volumes that need to be backed up using Restic. Every pod containing a volume to be backed up using Restic must be annotated with the volume’s name using the backup.velero.io/backup-volumes annotation.

Instructions to back up using this approach are as follows:

1. Run the following for each pod that contains a volume to back up:

   kubectl -n YOUR_POD_NAMESPACE annotate pod/YOUR_POD_NAME backup.velero.io/backup-volumes=YOUR_VOLUME_NAME_1,YOUR_VOLUME_NAME_2,...
   

   where the volume names are the names of the volumes in the pod spec.

   For example, for the following pod:

   apiVersion: v1
   kind: Pod
   metadata:
     name: sample
     namespace: foo
   spec:
     containers:
     - image: k8s.gcr.io/test-webserver
       name: test-webserver
       volumeMounts:
       - name: pvc-volume
         mountPath: /volume-1
       - name: emptydir-volume
         mountPath: /volume-2
     volumes:
     - name: pvc-volume
       persistentVolumeClaim:
         claimName: test-volume-claim
     - name: emptydir-volume
       emptyDir: {}
   

   You’d run:

   kubectl -n foo annotate pod/sample backup.velero.io/backup-volumes=pvc-volume,emptydir-volume
   

   This annotation can also be provided in a pod template spec if you use a controller to manage your pods.

2. Take a Velero backup:

   velero backup create NAME OPTIONS...
   

3. When the backup completes, view information about the backups:

   velero backup describe YOUR_BACKUP_NAME
   

   kubectl -n velero get podvolumebackups -l velero.io/backup-name=YOUR_BACKUP_NAME -o yaml
   

To restore

Regardless of how volumes are discovered for backup using Restic, the process of restoring remains the same.

1. Restore from your Velero backup:

   velero restore create --from-backup BACKUP_NAME OPTIONS...
   

2. When the restore completes, view information about your pod volume restores:

   velero restore describe YOUR_RESTORE_NAME
   

   kubectl -n velero get podvolumerestores -l velero.io/restore-name=YOUR_RESTORE_NAME -o yaml
   

Limitations

• hostPath volumes are not supported. Local persistent volumes are supported.
• Those of you familiar with restic may know that it encrypts all of its data. Velero uses a static, common encryption key for all Restic repositories it creates. This means that anyone who has access to your bucket can decrypt your Restic backup data. Make sure that you limit access to the Restic bucket appropriately.
• An incremental backup chain will be maintained across pod reschedules for PVCs. However, for pod volumes that are not PVCs, such as emptyDir volumes, when a pod is deleted/recreated (for example, by a ReplicaSet/Deployment), the next backup of those volumes will be full rather than incremental, because the pod volume’s lifecycle is assumed to be defined by its pod.
• Restic scans each file in a single thread. This means that large files (such as ones storing a database) will take a long time to scan for data deduplication, even if the actual difference is small.
• If you plan to use Velero’s Restic integration to backup 100GB of data or more, you may need to customize the resource limits to make sure backups complete successfully.
• Velero’s Restic integration backs up data from volumes by accessing the node’s filesystem, on which the pod is running. For this reason, Velero’s Restic integration can only backup volumes that are mounted by a pod and not directly from the PVC. For orphan PVC/PV pairs (without running pods), some Velero users overcame this limitation running a staging pod (i.e. a busybox or alpine container with an infinite sleep) to mount these PVC/PV pairs prior taking a Velero backup.

Customize Restore Helper Container

Velero uses a helper init container when performing a Restic restore. By default, the image for this container is velero/velero-restic-restore-helper:<VERSION>, where VERSION matches the version/tag of the main Velero image. You can customize the image that is used for this helper by creating a ConfigMap in the Velero namespace with the alternate image.

In addition, you can customize the resource requirements for the init container, should you need.

The ConfigMap must look like the following:

apiVersion: v1
kind: ConfigMap
metadata:
  # any name can be used; Velero uses the labels (below)
  # to identify it rather than the name
  name: restic-restore-action-config
  # must be in the velero namespace
  namespace: velero
  # the below labels should be used verbatim in your
  # ConfigMap.
  labels:
    # this value-less label identifies the ConfigMap as
    # config for a plugin (i.e. the built-in restic restore
    # item action plugin)
    velero.io/plugin-config: ""
    # this label identifies the name and kind of plugin
    # that this ConfigMap is for.
    velero.io/restic: RestoreItemAction
data:
  # The value for "image" can either include a tag or not;
  # if the tag is *not* included, the tag from the main Velero
  # image will automatically be used.
  image: myregistry.io/my-custom-helper-image[:OPTIONAL_TAG]

  # "cpuRequest" sets the request.cpu value on the restic init containers during restore.
  # If not set, it will default to "100m". A value of "0" is treated as unbounded.
  cpuRequest: 200m

  # "memRequest" sets the request.memory value on the restic init containers during restore.
  # If not set, it will default to "128Mi". A value of "0" is treated as unbounded.
  memRequest: 128Mi

  # "cpuLimit" sets the request.cpu value on the restic init containers during restore.
  # If not set, it will default to "100m". A value of "0" is treated as unbounded.
  cpuLimit: 200m

  # "memLimit" sets the request.memory value on the restic init containers during restore.
  # If not set, it will default to "128Mi". A value of "0" is treated as unbounded.
  memLimit: 128Mi

  # "secCtxRunAsUser" sets the securityContext.runAsUser value on the restic init containers during restore.
  secCtxRunAsUser: 1001

  # "secCtxRunAsGroup" sets the securityContext.runAsGroup value on the restic init containers during restore.
  secCtxRunAsGroup: 999

  # "secCtxAllowPrivilegeEscalation" sets the securityContext.allowPrivilegeEscalation value on the restic init containers during restore.
  secCtxAllowPrivilegeEscalation: false

  # "secCtx" sets the securityContext object value on the restic init containers during restore.
  # This key override  `secCtxRunAsUser`, `secCtxRunAsGroup`, `secCtxAllowPrivilegeEscalation` if `secCtx.runAsUser`, `secCtx.runAsGroup` or `secCtx.allowPrivilegeEscalation` are set.
  secCtx: |
    capabilities:
      drop:
      - ALL
      add: []
    allowPrivilegeEscalation: false
    readOnlyRootFilesystem: true
    runAsUser: 1001
    runAsGroup: 999

Troubleshooting

Run the following checks:

Are your Velero server and daemonset pods running?

kubectl get pods -n velero

Does your Restic repository exist, and is it ready?

velero restic repo get

velero restic repo get REPO_NAME -o yaml

Are there any errors in your Velero backup/restore?

velero backup describe BACKUP_NAME
velero backup logs BACKUP_NAME

velero restore describe RESTORE_NAME
velero restore logs RESTORE_NAME

What is the status of your pod volume backups/restores?

kubectl -n velero get podvolumebackups -l velero.io/backup-name=BACKUP_NAME -o yaml

kubectl -n velero get podvolumerestores -l velero.io/restore-name=RESTORE_NAME -o yaml

Is there any useful information in the Velero server or daemon pod logs?

kubectl -n velero logs deploy/velero
kubectl -n velero logs DAEMON_POD_NAME

NOTE: You can increase the verbosity of the pod logs by adding --log-level=debug as an argument to the container command in the deployment/daemonset pod template spec.

How backup and restore work with Restic

Velero has three custom resource definitions and associated controllers:

• ResticRepository - represents/manages the lifecycle of Velero’s restic repositories. Velero creates a Restic repository per namespace when the first Restic backup for a namespace is requested. The controller for this custom resource executes Restic repository lifecycle commands – restic init, restic check, and restic prune.

  You can see information about your Velero’s Restic repositories by running velero restic repo get.

• PodVolumeBackup - represents a Restic backup of a volume in a pod. The main Velero backup process creates one or more of these when it finds an annotated pod. Each node in the cluster runs a controller for this resource (in a daemonset) that handles the PodVolumeBackups for pods on that node. The controller executes restic backup commands to backup pod volume data.

• PodVolumeRestore - represents a Restic restore of a pod volume. The main Velero restore process creates one or more of these when it encounters a pod that has associated Restic backups. Each node in the cluster runs a controller for this resource (in the same daemonset as above) that handles the PodVolumeRestores for pods on that node. The controller executes restic restore commands to restore pod volume data.

Backup

1. Based on configuration, the main Velero backup process uses the opt-in or opt-out approach to check each pod that it’s backing up for the volumes to be backed up using Restic.
2. When found, Velero first ensures a Restic repository exists for the pod’s namespace, by:
   • checking if a ResticRepository custom resource already exists
   • if not, creating a new one, and waiting for the ResticRepository controller to init/check it
3. Velero then creates a PodVolumeBackup custom resource per volume listed in the pod annotation
4. The main Velero process now waits for the PodVolumeBackup resources to complete or fail
5. Meanwhile, each PodVolumeBackup is handled by the controller on the appropriate node, which:
   • has a hostPath volume mount of /var/lib/kubelet/pods to access the pod volume data
   • finds the pod volume’s subdirectory within the above volume
   • runs restic backup
   • updates the status of the custom resource to Completed or Failed
6. As each PodVolumeBackup finishes, the main Velero process adds it to the Velero backup in a file named <backup-name>-podvolumebackups.json.gz. This file gets uploaded to object storage alongside the backup tarball. It will be used for restores, as seen in the next section.

Restore

1. The main Velero restore process checks each existing PodVolumeBackup custom resource in the cluster to backup from.
2. For each PodVolumeBackup found, Velero first ensures a Restic repository exists for the pod’s namespace, by:
   • checking if a ResticRepository custom resource already exists
   • if not, creating a new one, and waiting for the ResticRepository controller to init/check it (note that in this case, the actual repository should already exist in object storage, so the Velero controller will simply check it for integrity)
3. Velero adds an init container to the pod, whose job is to wait for all Restic restores for the pod to complete (more on this shortly)
4. Velero creates the pod, with the added init container, by submitting it to the Kubernetes API. Then, the Kubernetes scheduler schedules this pod to a worker node, and the pod must be in a running state. If the pod fails to start for some reason (i.e. lack of cluster resources), the Restic restore will not be done.
5. Velero creates a PodVolumeRestore custom resource for each volume to be restored in the pod
6. The main Velero process now waits for each PodVolumeRestore resource to complete or fail
7. Meanwhile, each PodVolumeRestore is handled by the controller on the appropriate node, which:
   • has a hostPath volume mount of /var/lib/kubelet/pods to access the pod volume data
   • waits for the pod to be running the init container
   • finds the pod volume’s subdirectory within the above volume
   • runs restic restore
   • on success, writes a file into the pod volume, in a .velero subdirectory, whose name is the UID of the Velero restore that this pod volume restore is for
   • updates the status of the custom resource to Completed or Failed
8. The init container that was added to the pod is running a process that waits until it finds a file within each restored volume, under .velero, whose name is the UID of the Velero restore being run
9. Once all such files are found, the init container’s process terminates successfully and the pod moves on to running other init containers/the main containers.

Velero won’t restore a resource if a that resource is scaled to 0 and already exists in the cluster. If Velero restored the requested pods in this scenario, the Kubernetes reconciliation loops that manage resources would delete the running pods because its scaled to be 0. Velero will be able to restore once the resources is scaled up, and the pods are created and remain running.

3rd party controllers

Monitor backup annotation

Velero does not provide a mechanism to detect persistent volume claims that are missing the Restic backup annotation.

To solve this, a controller was written by Thomann Bits&Beats: velero-pvc-watcher