Potential RabbitMQ connection leak

[xaka]

Report

I've got ~600 scaled objects with similar rabbitmq config:

  triggers:
  - metadata:
      host: amqp://guest:[email protected]:5672
      mode: QueueLength
      protocol: amqp
      queueName: <unique-id>:queue
      value: "1"
      vhostName: /
    type: rabbitmq

Count TOP-10 connections by host

rabbitmq-0:/# rabbitmqctl list_connections | awk '
NR > 2 {count[$2]++}
END {
    for (host in count) {
        print host, count[host]
    }
}' | sort -k2,2nr | head -n 10
172.31.17.39 10369
172.31.26.39 1305
172.31.12.223 907
172.31.41.74 388
172.31.37.14 2
172.31.10.181 1
172.31.10.40 1
172.31.10.46 1
172.31.11.119 1
172.31.11.209 1

In this case 172.31.17.39 is keda operator pod and it opened 10k connections so far (5 days). The number is growing over time until rabbitmq gets stuck.

Expected Behavior

Number of connections is stable over time and doesn't grow

Actual Behavior

Connections are being opened over time, but never closed/disposed

Steps to Reproduce the Problem

N/A

Logs from KEDA operator

N/A

KEDA Version

2.15.1

Kubernetes Version

1.29

Platform

Amazon Web Services

Scaler Details

RabbitMQ

Anything else?

No response