FastStream NATS connection fails with internal Kubernetes DNS name but works with direct ClusterIP

[natenberenstein]

Hey guys, not sure if this is the right place to ask this question, but hopefully somebody would be able to clarify something for me.

Description

I'm experiencing a DNS resolution issue with FastStream connecting to NATS in Kubernetes.

Setup

• NATS JetStream cluster deployed in Kubernetes within the nats namespace (helm chart: https://github.com/nats-io/k8s/blob/nats-1.3.16/helm/charts/nats/values.yaml)
• FastAPI/Python application running inside the same cluster within the ai namespace
• Both are running on external systems (not local development)

The Issue

• The following connection works nats://<svc_clusterIP>:4222 (using direct service IP)
• Doesn't work: nats://nats-dev.nats.svc.cluster.local:4222 (using internal DNS name)

Code

# bootstrap.py
async def _init_broker() -> NatsBroker:
    from faststream.nats import NatsBroker
    
    logger.info(
        f"Attempting to connect to {settings.nats_url} with protocol {settings.nats_protocol}"
    )
    broker = NatsBroker(servers=[settings.nats_url], protocol=settings.nats_protocol)
    await broker.start()
    logger.info("NATS broker connected")
    return broker

Question

Given that, why would FastStream fail to connect when using the internal Kubernetes DNS name, but succeed when using the direct ClusterIP? Is there something specific about how FastStream or the underlying asyncio-nats library handles DNS resolution that I'm missing? Or is this more related to how Nats Jetstream is configured in the Kubernetes cluster?

Thanks in advance!