NATS JetStream scaler assume the cluster runs on kubernetes

[mfadhlika]

Report

Jetstream scaler assume the cluster run on the kubertenes cluster, which makes the scaler failed to find consumer leader monitoring url if that's not the case.

Expected Behavior

The scaler get correct leader monitoring node url even if NATS Jetstream cluster is not on kubernetes

Actual Behavior

The scaler assume first item when node url split by . is the node name and always have <node>.<monitoringURL> as URL which may not be case when NATS Jetstream run on multiple VMs.

Steps to Reproduce the Problem

1. Create NATS Jetstream cluster using VMs with load balancer
2. Use Jetstream scaler as documented
3. Get cluster information

❯ curl -s "http://load-balancer-ip:8222/varz" | jq .cluster
{
  "name": "nats",
  "addr": "172.0.0.2",
  "cluster_port": 4221,
  "auth_timeout": 2,
  "urls": [
    "172.0.0.3:4221", <-- currently 172 considered as node name
    "172.0.0.4:4221"
  ],
  "tls_timeout": 2
}

Logs from KEDA operator

Get "http://172.nats.company.internal:8222/jsz?acc=$G&consumers=true&config=true": dial tcp: lookup 172.nats.company.internal on 10.200.0.10:53: no such host

KEDA Version

2.9.1

Kubernetes Version

1.24

Platform

Google Cloud

Scaler Details

NATS Jetstream

Anything else?

No response

[JorTurFer]

Interesting.
Are you willing to contribute with the fix?
Isn't the machine FQDNs enough?

[zroubalik]

There's open PR for this already.

[mfadhlika]

@JorTurFer @zroubalik I opened the PR regarding this #4102

[JorTurFer]

f**k, sorry, I didn't see it, my fault :(