Create and publish an official Dekaf Helm chart

[mattsre]

Thank you for this project, the Dekaf UI for Pulsar is a great alternative to Pulsar manager! I'd like to use this project as a shared console for developers to explore Pulsar with. We deploy on Kubernetes, but do not use the official Pulsar Helm chart which you added Dekaf support to here - apache/pulsar-helm-chart#637

It would be nice if there was a dedicated Dekaf UI Helm chart maintained in this repository which we could pull and use directly. I'm open to submitting a PR for this change if you would accept it. The Github container registry supports OCI, so we could publish the Helm chart to GHCR on your repository.

[mattsre]

The upstream Apache Pulsar chart could also be updated to pull a versioned copy of the Dekaf chart as a dependency rather than having to vendor the deployment/service/pvc files

[visortelle]

@mattsre, thank you for the good words about the project.

Feel free to open the PR!
You can fully replace the /helm directory. I used this dir about 3 years ago for development purposes.

[visortelle]

@mattsre hi,
Did you a chance to start the work on this issue?

If not, I'll probably implement it by myself in a week or two.

[mattsre]

Hi sorry for the delay here, I have not started on this. I can get to it in the next few weekends if you don't get to it first

[visortelle]

@mattsre hi,
If you're sure you'll have time for this, I would be very appreciated if you'd implement it.
If not, please tell me and I'll implement it by myself.

Thanks.

[mattsre]

Hey @visortelle I've done some work on this and am testing the chart internally atm. Will publish a PR this weekend with the chart for your review!

[visortelle]

@mattsre hi,
I noticed you do some work on Helm chart in you Dekaf fork.

I'm not sure that we need autoscaling for Dekaf at this moment. I'll try to explain why.
If a user starts a consumer session, then pause it, then continue consuming messages, there is a chance that "continue consumer session" API request will be handled by another Dekaf instance. Dekaf stores information about current consumer sessions in-memory, and this way user may get an error.
It may be fixed by using Kubernetes sticky sessions, but it requires different set of annotations for different ingress controllers. Probably we can keep support for replicas > 1 and autoscaling in Helm chart, but in this case autoscaler should be disabled by default and we should add some comments in Helm values or/and README on how to deal with sticky sessions.