Backups
Because clusters host user services, many of which store data on physical disks, SHIPOPS provides backup system configuration.
Backup Replicas and Storage
The main backup settings are the number of replicas and the connection details for an external S3 service. The number of replicas is set automatically when the cluster is created, but it can be changed later if necessary.
If you create a cost-saving cluster with one system node, it always has one replica. If you create a fault-tolerant cluster, two replicas are configured by default, which is usually enough for fault tolerance. Additional replicas create extra copies of the data. With two replicas, one copy of the data is created; with three replicas, two copies are created. More than two replicas slows services down because of copying overhead.
As the external S3 service, you can use either the service provided by the same provider from which you rent the cluster servers, or any other S3 service. To reduce network latency, we recommend using an S3 service in the same region as the servers of the cluster that will use it. The connection requires a URL, bucket, key, and secret. For an example of obtaining these details from a provider, see Connecting a DigitalOcean S3 Service.
Backup Interface Access
After configuring the S3 connection, use the backup system web interface to create and restore backups. This is done through the Longhorn web interface, which can be enabled by configuring HTTP routing to the Volumes kube and selecting the longhorn-frontend service.
The Longhorn web interface does not have built-in authentication, so the following protection measures should be used:
- Do not bind it to a public domain name. Access to it should be only through VPN.
- Enable authentication with SHIPOPS tools by setting a username and password in the HTTP routing configuration form.