Migration Path for ZooKeeper-based Kafka Clusters without NodePools to KRaft

[artempol-dev]

Migration Path for ZooKeeper-based Kafka Clusters without NodePools to KRaft

Issue Description

I'm looking for guidance on migrating our ZooKeeper-based Kafka clusters to KRaft mode. Our situation differs from the official migration guide [strimzi.io/blog/2024/03/22/strimzi-kraft-migration/](https://strimzi.io/blog/2024/03/22/strimzi-kraft-migration/) in that we don't currently use NodePools at all.

Current Setup

• We have a standard ZooKeeper-based Kafka cluster (using kind: Kafka)
• We do not have any existing NodePool configurations
• We don't have any of these annotations on our resources:
  • strimzi.io/kraft
  • strimzi.io/node-pools
• Our high-availability configuration (such as pod affinity rules) is defined directly in the Kafka CR

Questions

1. What is the simplest migration path for clusters not using NodePools yet? Do we need to first migrate to NodePools before we can migrate to KRaft?

2. For our high-availability configuration (pod affinity, etc.) that is currently defined in the Kafka CR:

   • Should we move this configuration to the NodePool resources instead?
   • Will the Kafka resource stop creating brokers once NodePools are configured?

3. Is it better to create new cluster with nodepools first and migrate data with MM2 to it from source cluster ?

Thank you for your help and guidance!

[ppatierno]

The link you posted is not the "official migration guide", it's just a blog post showing the feature.
This is the official migration guide where you can find also the link to what to do if your ZooKeeper cluster is not using node pools yet: https://strimzi.io/docs/operators/latest/deploying#proc-deploy-migrate-kraft-str
I think the guide should answer all your questions above :-)

[artempol-dev]

@ppatierno Paolo, thank you!
So its needed to implement these steps:

1. To enable node-pool mode in my cluster and to deploy node-pool brokers is the first step
2. Migrating zookeeper to controller node-pool

Now I have only three questions:

1. Where to configure Highly availability config with nodepools? At the nodepools spec itself or in Kafka object as it was earlier ?
2. How to "say" Kafka object not to create brokers anymore as they are created by broker nodepool
3. bootstrap url stays the same?

[ppatierno]

Where to configure Highly availability config ... ?

What exactly are you referring to?

How to "say" Kafka object not to create brokers anymore as they are created by broker nodepool

When you enabled the strimzi.io/node-pools annotation on the Kafka resource, the operator will take care of it.

bootstrap url stays the same?

Absolutely, no changes from this pov.