[Bug]:ZooKeeper & AdminClient log DNS errors for *.svc hostnames (UnknownHostException on :9091 and <unresolved> :2181)

[Rodri2911]

Bug Description

When Strimzi 0.45.0 (ZooKeeper-based) runs on Kubernetes 1.32, the Cluster Operator, Topic Operator and User Operator log recurring DNS errors such as:

ERROR StaticHostProvider - Unable to resolve address: demo-cluster-zookeeper-0.demo-cluster-zookeeper-nodes.kafka.svc/<unresolved>:2181
WARN  NetworkClient [...] Error connecting to node demo-cluster-kafka-0.demo-cluster-kafka-brokers.kafka.svc:9091
java.net.UnknownHostException: demo-cluster-kafka-0.demo-cluster-kafka-brokers.kafka.svc

• The FQDN …svc.cluster.local does resolve instantly.
• Only the hostname without .cluster.local returns NXDOMAIN and triggers the warnings.
• This clutters logs and complicates alerting; the goal is to stop these messages from appearing.

Steps to reproduce

1. Deploy Strimzi 0.45.0 with the Helm chart and create a minimal ZooKeeper-based cluster (manifest below).
2. Wait until all pods are running; then check the logs of the Cluster Operator and Entity Operator.
3. Observe the repeated UnknownHostException / <unresolved> messages.

---

Strimzi version

0.45.0

Kafka version

3.9.0

Kubernetes version

v1.32.4+k0s.0

Installation method

Helm chart (strimzi/strimzi-kafka-operator) with:

extraEnvs:
  - name: KUBERNETES_SERVICE_DNS_DOMAIN
    value: cluster.local

Expected behavior

Operators should use hostnames that exist in CoreDNS (*.svc.cluster.local) or automatically append the suffix so that no DNS warnings appear in the logs.

Strimzi version

0.45.0

Kubernetes version

v1.32.4+k0s.0

Installation method

No response

Infrastructure

Bare-metal k0s cluster CoreDNS KubeRouter

Configuration files and logs

Configuration snippets

apiVersion: kafka.strimzi.io/v1beta2
kind: Kafka
metadata:
  name: demo-cluster
spec:
  kafka:
    version: 3.9.0
    replicas: 1
    listeners:
      - name: plain
        port: 9092
        type: internal
        tls: false
        configuration:
          useServiceDnsDomain: true
    storage:
      type: ephemeral
  zookeeper:
    replicas: 1
    storage:
      type: ephemeral
    config:
      quorumListenOnAllIPs: "true"
  entityOperator:
    topicOperator: {}
    userOperator: {}

---

Logs (examples)

Cluster Operator

ERROR StaticHostProvider - Unable to resolve address: demo-cluster-zookeeper-0.demo-cluster-zookeeper-nodes.kafka.svc/<unresolved>:2181
java.net.UnknownHostException: demo-cluster-zookeeper-0.demo-cluster-zookeeper-nodes.kafka.svc
        at java.net.InetAddress$CachedAddresses.get…

Topic Operator

WARN  NetworkClient:1082 - [AdminClient clientId=strimzi-topic-operator-…] Error connecting to node demo-cluster-kafka-0.demo-cluster-kafka-brokers.kafka.svc:9091
java.net.UnknownHostException: demo-cluster-kafka-0.demo-cluster-kafka-brokers.kafka.svc
        at java.net.InetAddress$CachedAddresses.get…

---

Additional context

• quorumListenOnAllIPs: "true" prevents ZooKeeper from hanging but does not silence the warnings.
• useServiceDnsDomain: true fixes listener 9092, not listener 9091 nor ZooKeeper server.X= entries.
• We cannot migrate to Strimzi 0.46+ because our workloads still rely on ZooKeeper, so we'd appreciate a solution or mitigation for 0.45.x clusters.
• Work-arounds like rewriting DNS in CoreDNS would affect the whole cluster; a Strimzi-side fix would be preferable.

[scholzj]

The domain name should resolve even without the cluster suffix. So this seems like an issue with your DNS.

[scholzj]

(And no, your steps to reproduce do not reproduce because it is related to your environment)