Automatic Etcd Backups and Restore

Through KKP you can set up automatic scheduled etcd backups for your user clusters, and easily restore etcd to its previous state.

Firstly, you need to enable and configure at least one backup destination (backup bucket, endpoint and credentials). To see how, check Etcd Backup Destination Settings.

It is recommended to enable EtcdLauncher on the clusters. It is required for the restore to work.

Etcd Backups

When the etcd backups are configured for a seed, project admins can create automatic Etcd Backups for the project clusters.

kind: EtcdBackupConfig
  name: daily-backup
  namespace: cluster-zvc78xnnz7
    kind: Cluster
    name: zvc78xnnz7
  schedule: '@every 20m'
  keep: 20
  destination: s3

Once created, the controller will start a new backup every time the schedule demands it (i.e. at 1:00 AM in the example), and delete backups once more than .spec.keep (20 in the example) have been created. The created backups are named based on the configuration name and the backup timestamp $backupconfigname-$timestamp. The backup is stored at the specified backup destination, which is configured per Seed.

The full specification for EtcdBackupConfig can be found in the CRD documentation. These resources can be either created via the UI (see below) or via kubectl directly.

It is also possible to do one-time backups (snapshots). The only change to the YAML example above for snapshots is omitting the schedule key.

Creating a Backup Schedule

EtcdBackups and Restores are project resources, and you can manage them in the Project view.

Etcd Backups View

To create a backup, just click on the Add Automatic Backup button. You have a choice of preset daily, weekly or monthly backups, or you can create a backup with a custom interval and keep time. The destination dropdown is based on the configured backup destinations for a given seed the cluster belongs to.

Etcd Backups Configuration

Listing Backups

To see what backups are available, click on a backup you are interested in, and you will see a list of completed backups.

Etcd Backups Details List

Alternatively, the list of backups is also published to the EtcdBackupConfig and can be fetched from the object’s status subresource via kubectl. It would look like this:

  - backupFinishedTime: "2021-03-16T01:48:22Z"
    backupName: zvc78xnnz7-jobtest1-2021-03-16T03:48:00
    backupPhase: Completed
    deleteFinishedTime: "2021-03-16T02:01:43Z"
    deleteJobName: zvc78xnnz7-backup-jobtest1-delete-s8sb8mdsp5
    deletePhase: Completed
    jobName: zvc78xnnz7-backup-jobtest1-create-rjjq9d4k6d
    scheduledTime: "2021-03-16T01:48:00Z"
  - backupFinishedTime: "2021-03-16T01:51:25Z"
    backupName: zvc78xnnz7-jobtest1-2021-03-16T03:51:00
    backupPhase: Completed
    deleteFinishedTime: "2021-03-16T02:04:08Z"
    deleteJobName: zvc78xnnz7-backup-jobtest1-delete-wcwvc8nmdt
    deletePhase: Completed
    jobName: zvc78xnnz7-backup-jobtest1-create-96tblznngk
    scheduledTime: "2021-03-16T01:51:00Z"
  - backupFinishedTime: "2021-03-16T01:54:21Z"
    backupName: zvc78xnnz7-jobtest1-2021-03-16T03:54:00
    backupPhase: Completed
    deleteFinishedTime: "2021-03-16T02:08:24Z"
    deleteJobName: zvc78xnnz7-backup-jobtest1-delete-cgmqhlw5fw
    deletePhase: Completed
    jobName: zvc78xnnz7-backup-jobtest1-create-v2gz2ntn46
    scheduledTime: "2021-03-16T01:54:00Z"

One-Time Backups

You can also create one-time backup snapshots, they are set up similarly to the automatic ones, with the difference that they do not have a schedule or keep count set.

Etcd Backup Snapshots

Restoring from Backup

To restore a cluster etcd from a backup, go into a backup details as shown above and select the backup from which you want to restore.

What will happen is that the cluster will get paused, etcd will get deleted, and then recreated from the backup. When it’s done, the cluster will unpause.

Keep in mind that this is an etcd backup and restore. The only thing that is restored is the etcd state, not application data or similar.

If you have running pods on the user cluster, which are not in the backup, it’s possible that they will get orphaned. Meaning that they will still run, even though etcd(and K8s) is not aware of them.

EtcdRestore can be marked with phase EtcdLauncherNotEnabled. You should enable EtcdLauncher on the cluster for restore to proceed.

Starting a Restore

If you want to restore from a specific backup, just click on the restore from backup icon.

Etcd Backups Restore button

Etcd Backups Cluster Restore

This will create an EtcdRestore object for your cluster. You can observe the progress in the Restore list.

Etcd Restore List

Starting Restore via kubectl

To restore a cluster from am existing backup via kubectl, you simply create a restore resource in the cluster namespace:

kind: EtcdRestore
  name: cluster-restore
  namespace: cluster-zvc78xnnz7
    kind: Cluster
    name: zvc78xnnz7
  backupName: daily-backup-2021-03-02T15:24:00
  destination: s3

This needs to reference the backup name from the list of backups (shown above).

Restore Progress

In the cluster view, you may notice that your cluster is in a Restoring state, and you can not interact with it until it is done.

Cluster Restoring

When it’s done, the cluster will get un-paused and un-blocked, so you can use it. The Etcd Restore will go into a Completed state.

Etcd Restore Completed


By default, KKP uses container snippets to run pre-defined scripts that upload backups to and deletes them from an S3-compatible storage.

They are configured in KubermaticConfiguration via:

  • .spec.seedController.backupDeleteContainer for backup deletion.
  • .spec.seedController.backupStoreContainer for backup file upload.

Their defaults can be reviewed in the reference KubermaticConfiguration. Both containers use the image d3fk/s3cmd to interact with S3-compatible storage. The configuration options mentioned above hold a full container specification, which means that a fullĀ container specification can be passed. This way uploads and deletions can be adjusted to individual needs. The dashboard and configuration objects expect the default containers to be capable of handling S3 endpoints.