Cluster Restore

In its default configuration, the Anvil Operator will back up your Anvil Cluster every 24 hours, with the internal databases backed up continuously. You can restore your cluster to any previous time by creating a ClusterRestore Resource. For example:

apiVersion: anvil.works/v1
kind: ClusterRestore
metadata:
  namespace: anvil
  name: maple-restore
spec:
  targetTime: "2025-04-28T12:00:00Z"
  source:
    pvc:
      claimName: anvil-maple-backup-b7fbaf

Configuration Reference

replaceClusterName

The name of the cluster to restore to. If not specified, the original name of the backed-up cluster will be used.

encryptionKeySecretName

The name of a Secret containing an encryption key to use for the restore. If not specified, the encryption key name from the backed-up cluster will be used.

suffixToken

The restored resource names have a suffix token to ensure they’re unique. If not specified, this will be the last 6 characters of the Restore resource’s UID.

imagePullSecretName

The name of a kubernetes.io/dockerconfigjson Secret to be used when pulling images from the Anvil container registry.

The default name can be overridden with the ANVIL_IMAGE_PULL_SECRET environment variable.

imagePrefix

If set, this overrides the default container image prefix for imagePrefixPublic and imagePrefixPrivate.

imagePrefixPublic

The prefix to use for public Anvil container images. If not specified, imagePrefix is used. If that isn’t set either, "anvil.works/public/" is used.

imagePrefixPrivate

The prefix to use for private Anvil container images. If not specified, imagePrefix is used. If that isn’t set either, "anvil.works/on-site/" is used.

platformServerVersionTag

The version of the temporary platform server container used to restore the app source code.

busyboxImage

The Busybox image to use for init containers.

sshServerImage

The OpenSSH server image to use for SSH servers.

targetTime

The restore point for the cluster. This can be a timezone-aware ISO 8601 timestamp, or latest_backup to restore the the most recent backup operation.

Since the database WALs are continuously archived between backups using PostgreSQL Continuous Archiving, you can restore to a more recent time than latest_backup by specifying the exact time to restore to.

defaultStorageClass

The default storage class to use for restored PVCs. If no storage class name is provided, the Kubernetes default will be used.

Restore Source

Configure the location of the backups to restore from. This is one of:

• s3: Restore from an S3 bucket. Currently this is only possible when running in an AWS EKS cluster.
• ssh: Restore from a remote server via SSH.
• nfs: Restore from an NFS volume in the cluster.
• pvc: Restore from a PVC in the cluster.

source.s3

source.s3.bucketName

The name of the S3 bucket to restore from.

source.s3.pathPrefix

The directory in the S3 bucket used to store the backup files.

source.ssh

source.ssh.hostname

The hostname of the SSH server to restore from.

source.ssh.port

The port of the SSH server.

source.ssh.directory

The directory on the remote server used to store the backup files.

source.ssh.username

The username to use for the SSH connection.

source.ssh.authSecretName

The name of a kubernetes.io/ssh-auth Secret containing an SSH private key to use for the restore.

source.ssh.knownHostsSecretName

The name of an Opaque Secret containing a known hosts file entry for the SSH server.

For example, you could create the following Secret and use it with knownHostsSecretName: my-known-hosts:

kubectl create -n anvil secret generic my-known-hosts --from-file="known_hosts"

source.nfs

source.nfs.server

The hostname of the NFS server.

source.nfs.export

The exported directory on the NFS server.

source.nfs.directory

The subdirectory inside the NFS export used to store the backup files.

source.pvc

source.pvc.claimName

The name of the PVC to use for the restore.

source.pvc.directory

The directory in the PVC used to store the backup files.

Pods

Configure the Kubernetes pods used to restore the cluster.

Each pod definition has a resources and affinity field. If neither of these are specified for a pod, the top-level pods.resources and pods.affinity will be used.

For example, the following specifies a 2GiB memory request for the containers in the restoreServer pod, but uses a 1GiB request for the containers in all other pods:

pods:
  resources:
    requests:
      memory: 1Gi
  restoreServer:
    resources:
      requests:
        memory: 2Gi

pods.resources

Configure the resources requests and limits for the containers in the pod. See the Kubernetes Resource Management documentation for more information.

pods.resources.requests

pods.resources.limits

pods.affinity

pods.affinity.node

Configure the node affinity for the pod by specifying a dictionary of keys and values to be used as requiredDuringSchedulingIgnoredDuringExecution.nodeSelectorTerms.

For example, the following specifies that the pod must be scheduled on anvil-node-0:

pods:
  affinity:
    node:
      kubernetes.io/hostname: anvil-node-0

See the Kubernetes Node Affinity documentation for more information.

pods.affinity.pod

Configure the pod affinity for the pod by specifying a dictionary of keys and values to be used as requiredDuringSchedulingIgnoredDuringExecution.labelSelectors.

For example, the following specifies that the pod must be scheduled on the same node as any pods with the label foo=bar:

pods:
  affinity:
    pod:
      foo: bar

See the Kubernetes Pod Affinity documentation for more information.

pods.restoreServer

Pod configuration (resources and affinity) for the SSH server used during the restore.

pods.gitRestore

Pod configuration (resources and affinity) for restoring the git repository.

pods.databaseRestore

Pod configuration (resources and affinity) for restoring the databases.

Do you still have questions?

Our Community Forum is full of helpful information and Anvil experts.