How do I back up and restore MongoDB with mongodump and mongorestore?

mongodump and mongorestore (MongoDB Database Tools, installed separately from the server in recent versions) are the simplest logical backup.

Dump one database, compressed, into a single archive

mongodump --uri="mongodb://backup:PASS@127.0.0.1:27017/app?authSource=admin" \
  --archive=/var/backups/mongo/app-$(date +%F).gz --gzip

Restore

mongorestore --uri="mongodb://admin:PASS@127.0.0.1:27017/?authSource=admin" \
  --archive=/var/backups/mongo/app-2026-11-09.gz --gzip --drop

--drop removes existing collections first so the restore matches the backup. To restore into a different database name use --nsFrom='app.*' --nsTo='app_copy.*'.

Consistency on replica sets. Add --oplog to a full-instance dump (not with --db) to capture writes made during the dump, and --oplogReplay on restore, so you get a point-in-time consistent result.

Backup user. Use the built-in backup role (and restore for restores) instead of root.

Automate with cron, rotate with find ... -mtime +14 -delete, and copy offsite just as with MySQL. Keep secrets out of the command line with a config file via --config (YAML with uri: ...).

Larger deployments

• Filesystem or LVM snapshots, or cloud snapshots, are faster for big data. For WiredTiger use journaling and snapshot the whole data directory.
• Managed Atlas clusters provide continuous cloud backups.

Pitfalls

• Copying /var/lib/mongodb while the server runs gives a corrupt backup.
• Indexes are rebuilt on restore, so time is dominated by index building.
• Test restores on a scratch instance, and check the document counts.