How to do nsqd load balancing with its sticky connections?

Question

I have a golang microservice project in kubernetes environment, that is running several nsqd-0, nsqd-1, ... pods, which are controlled by its statefulSet and grouped under nsqd service.

The problem is when I want to make a producer, i import github.com/nsqio/go-nsq and create it p := nsq.NewProducer('nsqd:4150', nsq.NewConfig())

But the connections are sticky (persistent) so what happens is service provides me ip adress of random pod (say it nsqd-2), and all traffic goes to it. others are unused at all

For now I am using awful workaround, every 5 seconds I p.Stop() current producer and reinitialize so it connects to a different pod. But how to load balance it normally? Is there any cloud-native simple way of doing it?

Answer 1

Another workaround is we can inititalize n times based on number of nsqd and put round robin logic behind it on client side