Docs Menu
Docs Home
/
MongoDB Ops Manager
/
Backup & Restore
/
Back Up

Backup Process

On this page

  • Backup Definition and Operational States
  • Backup Process Flows

Backups depend upon which version of MongoDB your database is compatible. This Feature Compatibility Version ranges from the current version to one version earlier. For MongoDB 4.2, the FCV can be 4.0 or 4.2.

The backup process takes a snapshot of the data directory at its scheduled snapshot intervals. This process copies the data files in a MongoDB deployment, sending them over the network via Ops Manager to your existing snapshot storage. Your deployment can still handle read and write operations during the copying process.

The backup process works in this manner regardless of how snapshots are stored.

Note

With the new Backup process, there are no longer initial syncs. As a result of not having initial syncs, Ops Manager can support a wider array of customers such as those heavily using renameCollection.

Once backup has started, Ops Manager backs up the data as a ongoing and continuous process. This process continues creating snapshots as long as the head database remains synchronized with the database.

This process works like replica set data synchronization.

The backup process:

  1. Performs an initial sync to back up all of your existing data in its current state. In sharded clusters, this occurs on each shard and on the config servers.

    Note

    Conditions or Actions that Restart Initial Sync

    During the initial sync process, certain actions or conditions can restart the initial sync process. Avoid the following actions and conditions:

    Actions to Avoid during Initial Sync:

    • Restarting, shutting down, or changing the version or FCV value of the source database.

    • Renaming the collection of the source database.

    • Changing the $out value in the Aggregation Pipeline of the source database.

    • Restarting or shutting down Ops Manager Application or Backup Daemon.

    • Restarting, shutting down, or upgrading the MongoDB Agent.

    Conditions to Avoid during Initial Sync:

    • Head Directory is full.

    • Network connectivity between Ops Manager components is unstable.

  2. Takes snapshots of the data directory in a deployment as often as your snapshot schedule specifies and then transfers the snapshots to a storage system.

    Note

    Sharded Clusters also can enable checkpoints to permit restores at points in time between snapshots. To learn how sharded clusters use checkpoints, see checkpoints.

    Important

    You may use checkpoints for clusters that run MongoDB with Feature Compatibility Version of 4.0 or earlier. Checkpoints were removed from MongoDB instances with FCV of 4.2 or later.

  3. Monitors the oplog constantly and adds new database operations to the latest backup to keep the local Ops Manager copy of the data current.

The backup process works in this manner regardless of how snapshots are stored.

Backup Definition and Operational States

Each backup is defined as a job. Each job defines how much and how often data is backed up. Backup jobs are defined on a per-project basis.

Operational States

The following table lists the states of a backup job:

State
Retain Old Snapshots
Create New Snapshots
Active
Yes
Yes
Stopped
Yes
No
Inactive
No
No
State
Retain Old Snapshots
Create New Snapshots
Apply Oplogs
Active
Yes
Yes
Yes
Stopped
Yes
No
No
Inactive
No
No
No

Change Operational States

Once backup jobs are active for a project, they run without further intervention until they are stopped or terminated. The operator can change the state of a backup in the following ways:

Initial State
Desired State
Method
Inactive
Active
Click Start.
Active
Stopped
Click Stop.
Stopped
Active
Click Restart.
Stopped
Inactive

Click Terminate.

WARNING: Terminate deletes all retained backups.

Initial State
Desired State
Method
Inactive
Active after Initial Sync
Click Start.
Active
Stopped
Click Stop.
Stopped
Active after Initial Sync
Click Restart.
Stopped
Inactive

Click Terminate.

WARNING: Terminate deletes all retained backups.

Important

You may receive a Backup requires a resync alert for your backup jobs. This may require you to Resync a Backup. This is not a different state, but a triggering of a new Backup Process Flow. Once Initial Sync completes, the backup job becomes Active again.

Backup Process Flows

Once created, a backup job goes through the following process flow:

Diagram showing the flow of data for Ops Manager's backup components when using snapshot storage.

  1. When the cluster is ready for its scheduled snapshot, it determines an optimal available node to take the snapshot. In most cases, the mongod determines lowest priority secondary member as the preferred snapshot node. Other metrics can factor into determining the preferred node, such as how current the secondary is with the primary and the previously chosen snapshot's member.

  2. Once the mongod process determines the origin node for the snapshot, the backup process opens a $backupCursor on the targeted node.

    The $backupCursor, a storage engine layer mechanism, allows the database files in storage to be copied in a consistent state while still accepting writes.

  3. The MongoDB Agent Backup function copies and processes these data files.

  4. The MongoDB Agent Backup function sends the data files to Ops Manager.

  5. The backup process collects and transfers these files to the snapshot store that you choose to store your backup. Depending upon which snapshot store you chose to store your snapshot, a snapshot can be written out as:

    1. Blocks to a blockstore. Binary chunks written to a MongoDB database on the Ops Manager host.

    2. Blocks to an AWS S3 bucket. The metadata for those blocks is written to a MongoDB database on the Ops Manager host.

    3. Snapshot files to a file system store.

Note

To learn more about the characteristics of each storage method, see Backup Configuration Options.

Initial Backup

Diagram showing the flow of data for Ops Manager's backup components when using snapshot storage.

  1. The Backup-enabled MongoDB Agent connects to, and authenticates with, the databases associated with the backup job.

  2. The initial sync begins and enters its starting phase. Initial sync is a transition state between Inactive and Active. Initial Sync goes through a series of phases that are displayed on the Backup page to show progress. Backup streams the existing data to Ops Manager in 10 MB compressed bundles of documents called slices. Backup creates slices at the point in time when the snapshot was created. Ops Manager captures data inserted to the instance once the snapshot starts separately.

  3. The transferring phase begins as the slices are streamed and stored in the Oplog Store temporarily on the Backup Daemon's behalf. The Backup Daemon service cannot dedicate itself to processing the large stream of initial sync slices at the expense of processing other backup jobs. The Oplog Store stores the slices until the Backup Daemon can fetch them. The Oplog Store is created when the first snapshot store is created.

  4. While Backup is streaming the data, it tails the oplog. This tailing collects any differences between the state of the deployment database when the backup began and the deployment database's current state. The oplog entries are sent in 10 MB compressed bundles of documents called oplog slices. These two streams of slices are collected in parallel to reduce the time needed to construct a complete snapshot.

  5. The building phase begins once Ops Manager receives the first batch of initial sync slices. In this phase, Ops Manager creates a local version of the backed up database called a head database on the host running the Backup Daemon service.

  6. Ops Manager uses the Backup Daemon service to insert the documents stored in the Oplog Store into the head database.

  7. The applying oplogs phase begins as Ops Manager applies the tailed oplog entries into the head database.

  8. During the fetching missing documents phase, Ops Manager queries the deployment database for documents missed during document insertion. Ops Manager inserts the missing documents found in the deployment database into the head database.

  9. After inserting the missing documents, the creating indexes phase begins as Ops Manager creates all of the indexes found in the deployment databases in the head database. When the indexes finish, the initial sync ends and the phase changes to complete.

  10. Depending upon which snapshot store you chose to store your snapshot, a snapshot can be written out as:

    1. Blocks to a blockstore.

    2. Blocks to an AWS S3 bucket. The metadata for those blocks is written to a MongoDB database on the Ops Manager host.

    3. Snapshot files to a file system store.

Note

The characteristics of each storage method is covered in Backup Configuration Options.

Subsequent Backups

The head database works as a full copy of the deployment database. It needs oplogs applied to it on a regular basis to keep its data synchronized with the deployment database. Snapshots are generated from the data stored in the head database according to your snapshot schedule.

Once the first full backup is completed, each active backup job follows this process:

  1. Backup tails the deployment's oplog.

  2. Backup routinely batches new oplog entries in oplog slices and transfers them to Ops Manager.

  3. Ops Manager stores the oplog entries in the Oplog Store.

  4. Ops Manager applies the new oplog entries from the oplog slices to the head database that stores the deployment backup.

  5. Ops Manager creates a new snapshot and stores it in the snapshot store as specified in your snapshot schedule.

Back

Back Up

Next

Preparations