Restore Overview


When you restore a deployment from a backup, you select a restore point and Cloud Manager provides you with one or more restore files. If you use Cloud Manager automation, Cloud Manager can automatically restore the files for you. Otherwise, you manually restore the files by copying them yourself to the servers you choose.

You can restore a single MongoDB process or an entire replica set or sharded cluster. You select whether to restore from an existing snapshot or a restore point between snapshots. For the latter, Cloud Manager builds a custom snapshot based on the restore point you choose. Installing from a custom snapshot takes longer as Cloud Manager must build the snapshot by applying the oplog to the previous regular snapshot. For sharded clusters, you must enable checkpoints before you can restore to a point between snapshots.

You can restore to either new hardware or existing hardware. If you use existing hardware and perform the restore manually, create a new path for the MongoDB data directory. Do not use existing data directories.

Automated Restore

If you choose to have Cloud Manager automatically restore your backup, Cloud Manager removes all existing data from the target servers and replaces that data with new backup data from your snapshot. For automated restores, you must have:

You can choose to restore to a cluster of a different group:

  • To restore to another Cloud Manager group, you must have Automation Admin or Backup Admin role for the destination group.
  • To restore to another MongoDB Atlas group, you must have Group Owner role for the destination group.


Under certain conditions, an automated restore fails because either:

If the backup and destination database storage engines or settings do not match, mongod cannot start once the backup is restored. At this time, you can either:

  • Not restore this snapshot, or
  • Change the storage engine or settings of the destination sharded cluster or replica set to match the configuration of the snapshot.

Manual Restore

For a manual restore, Cloud Manager provides each restore file as an archive (.tar or .tar.gz) containing the complete copy of the data directory.

  • For a replica set, Cloud Manager provides one archive that you copy to each replica set member.
  • For a sharded cluster, Cloud Manager provides one archive for the config servers and separate archives for each shard.

With a manual restore, Cloud Manager can provide the files in two ways:

Download via HTTP
Cloud Manager provides a link to download the restore files.
Transfer via secure copy (SCP)

Cloud Manager transfers the restore archive(s) or the individual data files to a target directory on a destination server of your choosing.


You need to generate a key pair before using SCP. SCP transfers files faster than HTTP.


Microsoft Windows does not include SCP. Installing SCP is outside the scope of this manual.

If you choose to receive archive files, you must extract the contents of each archive, then reconstruct your data.

This is generally faster than choosing to download the individual data files but requires sufficient space on the destination server for both the archive file and the extracted contents.