Nodeselector not working for Kafka and Zookeeper

[fayazahamedj]

Hello folks, I tried to implement NodeSelector for both Kafka & zookeeper as saw in one of the discussions, but it's not taking effect. Have shared the YAML snippet only for zookeeper as the complete YAML is lengthy. I tried nodeAffinity on the Kafka Strimzi Pod set & it is working fine, however when creating under Kafka cluster not working.

Have shared the complete Kafka YAML as attachment. Could you help on this??

apiVersion: kafka.strimzi.io/v1beta2
kind: Kafka
metadata:
 name: my-cluster
 namespace: streaming-telemetry
spec:
  zookeeper:
     replicas: 3
     storage:
       size: 2Gi
       type: persistent-claim
     template:
        pod:
           affinity:
              nodeAffinity:
                 requiredDuringSchedulingIgnoredDuringExecution:
                    nodeSelectorTerms:
                       - matchExpressions:
                              - key: nodetype
                                 operator: In
                                 values:
                                 - kafka

[fayazahamedj]

kafka.txt

[scholzj]

I'm not sure I follow what you mean with I tried nodeAffinity on the Kafka Strimzi Pod set & it is working fine, however when creating under Kafka cluster not working. -> can you elaborate on this, please?

[scholzj]

BTW: It looks like your cluster is using an old Kafka version:

  conditions:
    - lastTransitionTime: '2023-10-26T16:08:12.190040865Z'
      message: >-
        Unsupported Kafka.spec.kafka.version: 3.4.0. Supported versions are:
        [3.5.0, 3.5.1, 3.6.0]
      reason: UnsupportedKafkaVersionException
      status: 'True'
      type: NotReady

So if you change the affinity, it will not be applied because of this. You have to upgrade Kafka to fix this first. So maybe that is the problem?

[fayazahamedj]

@scholzj Yes, you are correct it is due to the old Kafka version. After instantiating a new Kafka instance, the nodeselector worked. I have one more doubt, is it possible to change the Kafka version for a running Kafka instance or do I have to create a fresh Kafka instance? And if able to change the Kafka version for a running Kafka instance, will changing it have any impact as such?

[scholzj]

The Kafka version is configgured in .spec.kafka.version. You can change it there and if possible, the operator will upgrade / downgrade your Kafka. So no need to deploy new cluster. However:

• It needs to be one of the supported Kafka versions, otherwise you get the error you got above
• For downgrades, it also depends on your configuran of inter.broker.protocol.version and some other factors.