Each Ark operation -- on-demand backup, scheduled backup, restore -- is a custom resource, defined with a Kubernetes [Custom Resource Definition (CRD)][20] and stored in [etcd][22]. Ark also includes controllers that process the custom resources to perform backups, restores, and all related operations.
Ark is ideal for the disaster recovery use case, as well as for snapshotting your application state, prior to performing system operations on your cluster (e.g. upgrades).
1. Uploads a tarball of copied Kubernetes objects into cloud object storage.
1. Calls the cloud provider API to make disk snapshots of persistent volumes, if specified.
You can optionally specify hooks to be executed during the backup. For example, you might
need to tell a database to flush its in-memory buffers to disk before taking a snapshot. [More about hooks][10].
Note that cluster backups are not strictly atomic. If Kubernetes objects are being created or edited at the time of backup, they might not be included in the backup. The odds of capturing inconsistent information are low, but it is possible.
The **schedule** operation allows you to back up your data at recurring intervals. The first backup is performed when the schedule is first created, and subsequent backups happen at the schedule's specified interval. These intervals are specified by a Cron expression.
Scheduled backups are saved with the name `<SCHEDULE NAME>-<TIMESTAMP>`, where `<TIMESTAMP>` is formatted as *YYYYMMDDhhmmss*.
The **restore** operation allows you to restore all of the objects and persistent volumes from a previously created backup. You can also restore only a filtered subset of objects and persistent volumes. Ark supports multiple namespace remapping--for example, in a single restore, objects in namespace "abc" can be recreated under namespace "def", and the objects in namespace "123" under "456".
The default name of a restore is `<BACKUP NAME>-<TIMESTAMP>`, where `<TIMESTAMP>` is formatted as *YYYYMMDDhhmmss*. You can also specify a custom name. A restored object also includes a label with key `ark.heptio.com/restore-name` and value `<RESTORE NAME>`.
By default, `ark backup create` makes disk snapshots of any persistent volumes. You can adjust the snapshots by specifying additional flags. Run `ark backup create --help` to see available flags. Snapshots can be disabled with the option `--snapshot-volumes=false`.
When you create a backup, you can specify a TTL by adding the flag `--ttl <DURATION>`. If Ark sees that an existing backup resource is expired, it removes:
Heptio Ark treats object storage as the source of truth. It continuously checks to see that the correct backup resources are always present. If there is a properly formatted backup file in the storage bucket, but no corresponding backup resource in the Kubernetes API, Ark synchronizes the information from object storage to Kubernetes.
Likewise, if a backup object exists in Kubernetes but not in object storage, it will be deleted from Kubernetes since the backup tarball no longer exists.
