[Bug]: ObservedGeneration in KafkaNodePool

[naveenthangaraj03]

Bug Description

• If we try to change the CPU, Memory, Volume or Replica in kafkanodepool, then the field observedGeneration will change once the controller initiates the changes

Steps to reproduce

No response

Expected behavior

• If we change the spec fields (CPU, Memory, Volume or Replica) in kafkanodepool then the observedGeneration field need to be changed after the controller finishes with the updates.

Strimzi version

0.46.0

Kubernetes version

Kubernetes 1.27

Installation method

Yaml files

Infrastructure

Bare-metal

Configuration files and logs

• Before updating the cpu and memory

status:
  clusterId: SvuTau6TS6SAr4ADyX5DXA
  conditions: []
  labelSelector: strimzi.io/cluster=kafka-fi,strimzi.io/name=kafka-fi-kafka,strimzi.io/kind=Kafka,strimzi.io/pool-name=kafka-fi-dual-role
  nodeIds:
  - 0
  - 1
  - 2
  - 3
  - 4
  observedGeneration: 5
  replicas: 5
  roles:
  - controller
  - broker

• After updating the cpu and memory

status:
  clusterId: SvuTau6TS6SAr4ADyX5DXA
  conditions: []
  labelSelector: strimzi.io/cluster=kafka-fi,strimzi.io/name=kafka-fi-kafka,strimzi.io/kind=Kafka,strimzi.io/pool-name=kafka-fi-dual-role
  nodeIds:
  - 0
  - 1
  - 2
  - 3
  - 4
  observedGeneration: 6
  replicas: 5
  roles:
  - controller
  - broker

• I had changed the cpu and memory for kafkanodepool but observedgeneration was updated once I initiated the update, it doesn't waited for the controller to finish the update, you can refer through the above image in that controller started to recreate the pod according to the update but only two pods are done but the observedgeneration is updated it doesn't waited for controller to finishes the update.

Additional context

No response

[im-konge]

Sorry, what is exactly the issue you hit?

[naveenthangaraj03]

So, you are trying to say that we can only update the observedGeneration in kafka cr after the update is done, but we can't add any condition or something when the update is in progress.

[scholzj]

I'm not sure what do you mean with we can.

But in Strimzi, the conditions describe the observed cluster state at the end of the reconciliation.

[naveenthangaraj03]

• What am trying to say is only you have idea to update the observedGeneration in kafka cr after any update is done?
• I will take a look into controller and if possible can I update the controller to add the condition when the reconciliation starts as well.

[scholzj]

• What am trying to say is only you have idea to update the observedGeneration in kafka cr after any update is done?

I might have some ideas how to deal with it. But as I said, the issue ultimately needs a proposal to be written and go through the approval process. So any ideas are just that - ideas.

• I will take a look into controller and if possible can I update the controller to add the condition when the reconciliation starts as well.

I'm not sure where you want to look into it or what controller you are talking about. But if you want to do it in Strimzi, it would definitely need to go through the proposal process. And speaking for my self, I also don't think we want any condition at the start of the reconciliation as it makes little sense and would just increase the load on the Kube APIs.

[naveenthangaraj03]

I understand now where you are coming from, let me go through the code of strimzi operator and propose an idea around it.