Etcd backups (assuming etcd backup is enabled) are currently stored on the control/master nodes only. If the control/master nodes go away, the etcd backups are lost.

Additionally, backups aren't synced between these nodes – thus, not all backups may be available, even if just one node is lost (assuming a multi-master cluster).

The UI/API should present options to store etcd backups on a Persistent Volume (PV) or another external storage location (e.g. Amazon S3, Backblaze B2, etc.) and/or ensure that the location specified for the etcd backup is a shared filepath (i.e. NFS).