Rails Mongoid仅配置单主机时能否自动识别MongoDB副本集主节点？

Great question—let's break this down clearly based on how MongoDB's Ruby driver (which Mongoid relies on) interacts with replica sets.

Why You're Seeing Traffic to Node1's Mongo From Node2 #

First, the reason Node2's web service is using Node1's Mongo even though it's not in mongoid.yml is because the MongoDB Ruby driver automatically discovers the full replica set topology after connecting to any single valid member. When your Node2 Rails app connects to its local mongo:27017 container, that Mongo node shares the complete replica set metadata (including Node1's Mongo and the arbiter) with the driver. The driver then uses this metadata to route requests appropriately—for example, sending writes to the primary node, and reads to whichever nodes match your read preference setting.

Can You Keep the Current Minimal Configuration? #

Technically, yes—your setup will continue to work because the driver handles automatic topology discovery. However, this is not recommended for production environments for a few key reasons:

• Initial connection risk: If the single configured node is down when your Rails app starts up, the app won't be able to connect to the replica set at all. There's no fallback to other healthy nodes in the set.
• Topology refresh delays: While the driver periodically refreshes its view of the replica set, if your initial configured node goes offline after startup, the driver may take time to detect the new primary or healthy secondary nodes, leading to temporary connection issues.
• Unpredictable failover behavior: Without a full list of nodes, the driver has fewer options to fall back on during failover events, increasing the chance of application downtime.

Best Practice: Configure All Replica Set Members #

For reliability and compliance with MongoDB/Mongoid best practices, you should update mongoid.yml on both nodes to include all replica set members. This ensures your app can connect to the replica set even if one or more nodes are down at startup, and speeds up topology discovery.

Here's what a proper replica set configuration looks like in mongoid.yml:

production:
  clients:
    default:
      database: your_app_db
      hosts:
        - node1-mongo:27017  # Replace with the actual reachable address of Node1's Mongo
        - node2-mongo:27017  # Replace with the actual reachable address of Node2's Mongo
        - arbiter-node:27017 # Replace with the arbiter's address
      options:
        replica_set: your_replica_set_name  # Match your replica set's name
        read_preference: primary_preferred  # Adjust based on your read needs

Key Note on Read/Write Operations #

Since you mentioned your workload includes both reads and writes:

• Writes are always automatically routed to the primary node by the driver, regardless of your initial configuration.
• Reads can be configured to go to secondary nodes (via read_preference settings like secondaryPreferred) once the driver has discovered the full topology. This works even with a single initial node, but again, having all nodes in the configuration makes this behavior more reliable.

内容的提问来源于stack exchange，提问作者dsanchez