Queue Groups in NATS: Load Balancing for Subscribers

In the previous articles, we looked at Publish–Subscribe and Request–Reply patterns in NATS. Pub/Sub is great for broadcasting messages to many subscribers, while Request–Reply is perfect for one-to-one communication. But what if you want to distribute messages across a group of workers so that only one of them processes each message?

This is where Queue Groups come in.

What are Queue Groups?

A Queue Group is a way to make multiple subscribers share the work for the same subject. Instead of all subscribers getting every message, each message is delivered to only one member of the group.

• Publishers don’t need to know about the group.
• Subscribers just join the same queue group by giving it a name.
• NATS automatically balances the messages among them.

Example in Go

Publisher:

nc.Publish("jobs", []byte("process this task"))

Subscribers in a Queue Group:

// Worker A
nc.QueueSubscribe("jobs", "workers", func(msg *nats.Msg) {
    fmt.Println("Worker A got:", string(msg.Data))
})

// Worker B
nc.QueueSubscribe("jobs", "workers", func(msg *nats.Msg) {
    fmt.Println("Worker B got:", string(msg.Data))
})

If the publisher sends 10 messages to jobs, NATS will spread them between Worker A and Worker B. Each message will only be delivered once to one worker.

Why Use Queue Groups?

1. Load Balancing
   • Multiple workers can share the load of incoming tasks.
   • Example: background jobs, image processing, or notification sending.
2. High Availability
   • If one worker goes down, the others in the group continue processing.
   • This reduces downtime and increases reliability.
3. Horizontal Scaling
   • You can add more subscribers to the queue group to handle higher traffic.
   • No need to change publishers — they keep sending to the same subject.