Skip to main content

Velero

Setting up Velero​

Configuration​

  • Plugin usage:
    • Azure: velero-plugin-for-microsoft-azure
    • AWS: velero-plugin-for-aws
  • Snapshot mechanism:
    • Azure: CSI snapshots via Azure Disk CSI Driver
    • AWS: EBS snapshots via AWS EBS CSI Driver
  • Storage bucket for Velero metadata:
    • Azure: Azure Blob Storage
    • AWS: Amazon S3
  • Backup naming convention: daily-backup-<namespace>-<timestamp>

Scheduling​

  • Daily full-volume backups via velero schedule CRDs
  • Retention policies configured using TTL settings in Velero schedules

Example value files can be found here.

How to check the backup in Velero​

  • using kubectl
kubectl get backup -n velero
  • using velero
velero get backups

Check backup schedules​

  • using kubectl
kubectl get schedules -n velero
  • using velero
velero get schedules

Triggering manual backups​

  • To trigger a manual backup using existing schedule
velero create backup manual-backup --from-schedule daily-backup`

Restoring using existing backups​

  • To restore from a backup
velero restore create --from-backup manual-backup
  • To verify restore request has been created
velero get restores

Docs​

Upgrade​

Velero can be updated by updating the helm chart.

NOTE: Please look at the respective plugin you use in velero, before updating the velero.

  • Version Matrix
ProviderLinks
AWShttps://github.com/vmware-tanzu/velero-plugin-for-aws#compatibility
CSIhttps://github.com/vmware-tanzu/velero-plugin-for-csi#compatibility
Azurehttps://github.com/vmware-tanzu/velero-plugin-for-microsoft-azure#compatibility

Issues​

Velero keeps a track of schedules even when they are deleted. https://github.com/vmware-tanzu/velero/issues/1333

It keeps publishing the metrics of older schedules even after removing them. The available solution for now is to kill the pods when no backups/restores are being performed.

Troubleshooting​

Install the Velero CLI. Use this command to describe and check logs of a particular backup.

# Get info about a backup
velero backup describe <backup_name>

# Get logs about a backup
velero backup logs <backup_name>

Velero uses snapshot-controller to backup PVCs by taking a Volume Snapshot. It uses the Snapshot APIs (snapshot.storage.k8s.io) to create a VolumeSnapshot and a VolumeSnapshotContent. The snapshot will be persisted to the cloud provider. E.g.- EBS Snapshots if you are using AWS.

Sometimes, the volumesnapshot objects remain in the cluster even after the backup has been deleted. We need to keep an eye on these objects taking up space and remove them to avoid huge cloud bills.

To delete a VolumeSnapshot, set the underlying VolumeSnapshotContent object's spec.deletionPolicy from Retain to Delete. Then delete the VolumeSnapshot and it will trigger the deletion of VolumeSnapshotContent as well as the Snapshot on your cloud provider.

Create SealedSecret for Velero using file​

kubectl create secret generic cloud-credentials -n velero  --from-file=cloud=cloud-credentials.yaml -o yaml --dry-run=client | kubeseal --controller-namespace system --controller-name sealed-secrets --format yaml > cloud.yaml

Where, contents of file cloud-credentials.yaml is shown below. From cloud to cloud the variable differs. And NOTE that this is only valid for Azure cloud.

AZURE_SUBSCRIPTION_ID=fkfkfdfdfdkfjlss
AZURE_TENANT_ID=wellllcmdndkdkd
AZURE_CLIENT_ID=3920keeiwid93ri3jm3idn3din3
AZURE_CLIENT_SECRET=ekji2ieje8eje3ij3i
AZURE_RESOURCE_GROUP=MC_k8s-prod-ddkkwji
AZURE_CLOUD_NAME=AwsPublicCloud

Create SealedSecret for Velero​

Depending on the cloud provider, run the script located inside /bin/<cloud> directory. (Currently only Azure is supported)

# help
./azure --help

# running the script
./azure --client-secret xxxx --cluster-name cluster.company.com

Script will generate credentials-velero.json in same directory. User has to apply that file to the desired cluster.

Create Token​

Generating a azure token with role as contributor, scope as a subscription id for lifelong.

az ad sp create-for-rbac --role="Contributor" --scopes="/subscriptions/<subscription-id>" --years 4000 --output json

References​