如何配置MongoDB副本集以仅复制指定集合或数据库？

Hey there! Great question—MongoDB's default setup replicates every database and collection in a replica set, but you can absolutely lock this down to only sync specific databases or collections using partial replication. Let's walk through exactly how to do this, step by step.

• Make sure you already have a basic replica set up (minimum 3 nodes: 1 primary, 2 secondaries/arbiter). All nodes need to run the same MongoDB version to avoid compatibility issues.
• Critical note: Nodes configured for partial replication cannot become primary nodes. They have to stay as secondaries, so we'll set their priority/vote count to 0 later to prevent election attempts.

This method lets a secondary node sync only the databases you explicitly list.

Step-by-Step: #

1. Stop the target secondary node's MongoDB service:

   sudo systemctl stop mongod
   

   (Use your custom stop command if you're not using systemd.)

2. Edit the node's config file (usually /etc/mongod.conf). In the replication section, add the only parameter to list your target databases:

   replication:
     replSetName: "myReplicaSet"  # Match your existing replica set name
     only: ["inventory", "sales"]  # List databases you want to replicate; single DB uses ["inventory"]
   

   Heads up: The only flag only works on secondary nodes—never set this on your primary.

3. Warning: Wipe the node's existing data directory (old full-replica data will conflict with filtered replication):

   sudo rm -rf /var/lib/mongodb/*  # Replace with your actual data path—double-check this!
   

4. Start the MongoDB service back up:

   sudo systemctl start mongod
   

5. Re-add the node to the replica set (with non-electable settings):
   Connect to your primary node via mongosh:

   mongosh --host <primary-ip>:27017
   

   Then run these commands:

   // Remove the old entry for the secondary
   rs.remove("<secondary-ip>:27017")
   // Re-add it with priority/vote set to 0 (non-electable)
   rs.add({
     _id: 1,  // Match the node's original _id in rs.conf()
     host: "<secondary-ip>:27017",
     priority: 0,
     votes: 0
   })
   // Verify the updated config
   rs.conf()
   

   The secondary will now only sync the databases you listed.

For finer control (syncing individual collections within a database), use the filter parameter with a namespace match.

Step-by-Step: #

1. Stop the target secondary node's MongoDB service (same as before):

   sudo systemctl stop mongod
   

2. Edit the config file to add the filter rule. For example, to sync only inventory.products and sales.orders:

   replication:
     replSetName: "myReplicaSet"
     filter: '{ "ns": { "$in": ["inventory.products", "sales.orders"] } }'
   

   • ns stands for "namespace" (format: database.collection)
   • Use $in to list multiple collections you want to sync

3. Wipe the node's data directory (same warning as before):

   sudo rm -rf /var/lib/mongodb/*
   

4. Start the service and re-add the node to the replica set with non-electable settings—follow the same steps as the database-level filtering section.

Connect to the filtered secondary via mongosh and run these checks:

// Check replication status
rs.status()
// List databases—only your target ones should appear
show dbs
// Check collections in the target DB
use inventory
show collections
// Confirm data matches the primary
db.products.find().count()

If the counts match the primary's collections, you're good to go!

• Partial replication nodes can never be primary—always set priority: 0 and votes: 0 to avoid cluster issues.
• Once you set only or filter, you can't switch back to full replication without wiping the node's data and reconfiguring.
• You can set up multiple secondaries with different filter rules if you need different subsets of data.
• Keep all nodes on the same MongoDB version—partial replication has strict compatibility requirements.

内容的提问来源于stack exchange，提问作者mohamed hamada