Backup operations use Cosmos DB change feed APIs to fetch the change feed of the collections. The change feed is used as follows:
-
For a full backup, the system fetches the change feed from the beginning of the collection.
-
For an incremental backup, the system fetches the changes from the last backup time.
-
The backup size of a collection is the total size of the change feed documents fetched for that collection during the backup operation.
-
The change feed read API does not exit until new documents are written. This means that backups might take longer to complete than you expect.
The backup operation retrieves the data from each collection and, if it fails, it is restartable from the point of failure. That is, the backup resumes from the last successfully fetched document from that collection.
To back up multiple collections at the same time, you can configure the number of streams at the collection group level. To provide backup load distribution and node failover functionality, you can configure multiple access nodes in the cloud instance.