For AI agents: a documentation index is available at https://www.mongodb.com/docs/llms.txt — markdown versions of all pages are available by appending .md to any URL path.

Daemons

This page lists all active Backup Daemons and provides configuration options. Ops Manager automatically detects Backup Daemons and displays them here. You can reconfigure daemons from this page. Changes can take up to 5 minutes to take effect.

Click Pre-Configure New Daemon at the bottom of the page if you want to add a daemon but do not want it to take new jobs. Type the <machine>:/<roothead path> in the text field above Pre-Configure New Daemon. If you omit the trailing slash, Ops Manager adds it automatically and always stores the directory in its canonical form, with a trailing slash. Click the button to configure the new Backup Daemon.

For each daemon, the page lists the server name, configuration, head space used, head space free, the number of replica sets backed up, the percentage of time the Backup Daemon Service was busy, and job runtime counts by 1 minute, 10 minutes, 60 minutes, less than or equal to 180 minutes, and greater than 180 minutes.

The page includes the following fields and links to manage and configure daemons:

Item
Action

Show Detailed JSON

Click the Show Detailed JSON link to display the Backup Daemon JSON. When the JSON displays, the View raw runtime data link appears following the code to view the raw data.

Move all heads

Click the Move all heads link to move all the Backup Daemon jobs that live on this daemon to a new daemon location.

  1. Click the link.

  2. Click the location.

  3. Click Move all heads.

Delete Daemon

Click the Delete Daemon link to delete the Backup Daemon.

You must restart the daemon to reconfigure the daemon after deletion.

If backup jobs are still bound to the daemon, Ops Manager denies the deletion and returns a BACKUP_CANNOT_REMOVE_DAEMON_CONFIG error. The check finds bound jobs regardless of how you entered the head directory, with or without a trailing slash. Ops Manager stores every head directory in its canonical form and matches jobs against that form. If you delete a daemon configuration that doesn't exist, Ops Manager returns a 404 error.

To decommission a daemon, first click Move all heads to move its jobs to another daemon, then delete the daemon. Move all heads reports how many jobs it moved.

Assignment

Select this to allow the daemon to process more jobs. If the daemon's disk fills, clear this so the host is not overloaded.

Backup Jobs

Select this to allow the daemon to process more backup jobs. Clear this when performing maintenance on a daemon's host.

Queryable Restore Jobs

Select this to allow the daemon to process queryable backups.

For the fastest possible restore time, configure a dedicated queryable restore daemon by selecting Queryable Restore Jobs and clearing the checkboxes for the other daemon roles, for example, Garbage Collection and Resource Usage.

Resource Usage

Select this to collect data for this daemon. This includes the following data for each backup job run on this daemon:

  • project

  • storage that head database uses

  • storage that the snapshot store uses

  • oplog operations

Navigate to the Resource Usage page to review the collected data.

Garbage Collection

Select this to allow the daemon to schedule groom jobs to remove unused blocks and expired snapshots.

Head Disk Type

From this menu, select SSD or HDD storage.

If you select SSD, Ops Manager sends jobs to this daemon only when:

  • The job generates a volume of oplog data that meets or exceeds a threshold of 1 GBph, or

  • Ops Manager has no available HDD daemon to handle the job.

Deployment Id

Unique identifier of the Deployment Region in which to run the daemon.

Assignment Labels

Type one or more labels that can be used to assign the daemon to a specific project.

Number of Workers

Type the number of concurrent backup tasks this daemon should process at one time.

To query a snapshot of a sharded cluster, the Backup Daemon requires at least one worker for the config server, one worker for each shard, and one worker for the mongos instance.

To query a snapshot of a replica set, the Backup Daemon requires at least one worker for the replica set.

For example, if you restore a queryable backup from a 3-shard cluster with 1 shard router (mongos), you need this value to be at least 5:

  • 1 per shard (3) +

  • 1 for the config server (1) +

  • 1 for the mongos

When the queryable backup begins, the Backup Daemon spins up 5 or more workers to manage these components.

Ops Manager stores each Backup Daemon head directory in a single canonical form: the path with one trailing slash. The API and UI accept a head directory with or without a trailing slash and always store and return the canonical form. For example, if you enter /data or /data//, Ops Manager stores and returns /data/.

When you upgrade to Ops Manager 8.0.28, 9.0.2, or a later version, a one-time migration canonicalizes the existing daemon data. The migration runs at startup, before Ops Manager serves requests, and performs the following actions:

  • Converts slash-less and multi-slash head directories to the canonical spelling, rebinds their jobs, and rewrites the group and restore daemonFilter entries.

  • Merges configurations that spelled the same identity in two ways. The canonical-spelling document wins, and the migration removes the duplicate and drops its settings, such as workers and assignment flags. The migration logs the full duplicate document so that you can reapply its settings by hand.

  • Skips any document that it can't migrate, such as a document with no hostname or with malformed machine or head fields, and logs an ERROR line in logs/mms-migration.log that names the document's _id. The skipped document is inert and Ops Manager starts normally. Repair the document instead of rolling back: delete it or restore its hostname:

    db.getSiblingDB("backupconfig").daemon.deleteOne({ _id: <_id from the log line>})db.getSiblingDB("backupconfig").daemon.updateOne( { _id: <_id from the log line> }, { $set: { "machine.machine": "<hostname>" } })

If the migration itself fails, for example because of an unexpected database error, startup aborts with a migrations failed message and Ops Manager doesn't start. The log names the problem. Fix the problem and start Ops Manager again instead of rolling back. The migration re-runs on the next start. To trigger a re-run, start Ops Manager with -Dforce-rerun=true. Do not add a migration. prefix to the flag name.

During a rolling upgrade, while any node still runs an earlier version, all API writes return a 405 WRITES_NOT_ALLOWED_DURING_APP_UPGRADES error. This behavior is by design and lifts when the last node completes its upgrade.

After the upgrade, you might observe the following effects:

  • Backup pickup can lag up to one hour for jobs that were in flight during the upgrade. This gap is self-healing and causes no data loss.

  • Head directory paths display with a trailing slash. Statistics rows stored under the old spelling become unreachable and age out on their own.

  • The migration logs to logs/mms-migration.log, which the group diagnostic archive does not collect. When you triage an upgrade, collect this file from the host.