Version:

Back up Kyma

Context

The Kyma cluster load consists of Kubernetes objects and volumes.

Object backup

Kyma relies on a managed Kubernetes cluster for periodic backups of Kubernetes objects to avoid any manual steps.

CAUTION: Automatic backup doesn't include Kubernetes volumes. Back up your volumes periodically either on demand, or set up a periodic job.

For example, Gardener uses etcd as the Kubernetes backing store for all cluster data. Gardener runs periodic jobs to take major and minor snapshots of the etcd database to include Kubernetes objects in the backup.

The major snapshot that includes all resources is taken on a daily basis, while minor snapshots are taken every five minutes.

If the etcd database experiences any problems, Gardener automatically restores the Kubernetes cluster using the most recent snapshot.

Volume backup

We recommend that you back up your volumes periodically with the VolumeSnapshot API resource, which is provided by Kubernetes. You can use your snapshot to provision a new volume prepopulated with the snapshot data, or restore the existing volume to the state represented by the snapshot.

Taking volume snapshots is possible thanks to Container Storage Interface (CSI) drivers, which allow third-party storage providers to expose storage systems in Kubernetes. For details on available drivers, see the full list of drivers.

You can create on-demand volume snapshots manually, or set up a periodic job that takes automatic snapshots periodically.

Back up resources using Velero

You can back up and restore individual resources manually or automatically with Velero. For more information, read the Velero documentation. Be aware that a full backup of a Kyma cluster isn't supported. Start with the existing Kyma installation and restore specific resources individually.

Create on-demand volume snapshots

If you want to provision a new volume or restore the existing one, create on-demand volume snapshots:

Gardener
AKS
GKE

Create a periodic snapshot job

You can also create a CronJob to handle taking volume snapshots periodically. A sample CronJob definition that includes the required ServiceAccount and roles looks as follows:

Click to copy

---
apiVersion: v1
kind: ServiceAccount
metadata:
  name: volume-snapshotter
---
apiVersion: rbac.authorization.k8s.io/v1
kind: Role
metadata:
  name: volume-snapshotter
  namespace: {NAMESPACE}
rules:
- apiGroups: ["snapshot.storage.k8s.io"]
  resources: ["volumesnapshots"]
  verbs: ["create", "get", "list", "delete"]
---
apiVersion: rbac.authorization.k8s.io/v1
kind: RoleBinding
metadata:
  name: volume-snapshotter
  namespace: {NAMESPACE}
roleRef:
  apiGroup: rbac.authorization.k8s.io
  kind: Role
  name: volume-snapshotter
subjects:
- kind: ServiceAccount
  name: volume-snapshotter
---
apiVersion: batch/v1beta1
kind: CronJob
metadata:
  name: volume-snapshotter
  namespace: {NAMESPACE}
spec:
  schedule: "@hourly" #Run once an hour, beginning of hour
  jobTemplate:
    spec:
      template:
        spec:
          serviceAccountName: volume-snapshotter
          containers:
          - name: job
            image: eu.gcr.io/kyma-project/tpi/k8s-tools:v20230110-ce11007d
            command:
              - /bin/bash
              - -c
              - |
                # Create volume snapshot with random name.
                RANDOM_ID=$(openssl rand -hex 4)
                cat <<EOF | kubectl apply -f -
                apiVersion: snapshot.storage.k8s.io/v1beta1
                kind: VolumeSnapshot
                metadata:
                  name: volume-snapshot-${RANDOM_ID}
                  namespace: {NAMESPACE}
                  labels:
                    "job": "volume-snapshotter"
                    "name": "volume-snapshot-${RANDOM_ID}"
                spec:
                  volumeSnapshotClassName: {SNAPSHOT_CLASS_NAME}
                  source:
                    persistentVolumeClaimName: {PVC_NAME}
                EOF
                # Wait until volume snapshot is ready to use.
                attempts=3
                retryTimeInSec="30"
                for ((i=1; i<=attempts; i++)); do
                    STATUS=$(kubectl get volumesnapshot volume-snapshot-${RANDOM_ID} -n {NAMESPACE} -o jsonpath='{.status.readyToUse}')
                    if [ "${STATUS}" == "true" ]; then
                        echo "Volume snapshot is ready to use."
                        break
                    fi
                    if [[ "${i}" -lt "${attempts}" ]]; then
                        echo "Volume snapshot is not yet ready to use, let's wait ${retryTimeInSec} seconds and retry. Attempts ${i} of ${attempts}."
                    else
                        echo "Volume snapshot is still not ready to use after ${attempts} attempts, giving up."
                        exit 1
                    fi
                    sleep ${retryTimeInSec}
                done
                # Delete old volume snapshots.
                kubectl delete volumesnapshot -n {NAMESPACE} -l job=volume-snapshotter,name!=volume-snapshot-${RANDOM_ID}