Skipper: Skipper Helm Chart

[kolbis]

Is there anyone which already created a production ready helm chart for skipper?
If so, can you please share it?

[szuecs]

We I don’t like helm and probably all of them in the internet are outdated.

https://github.com/zalando-incubator/kubernetes-on-aws/tree/dev/cluster/manifests/skipper has all manifests that you need to build a production setup. It’s used in >140 clusters.

[kolbis]

I appreciate the answer.
Just out of curiosity, why dont you like helm?

[szuecs]

Tiller is known to be unsafe.
I don’t like, if I have a templated mess that I can’t easily understand. It feels a bit like the puppet module forge.

In my opinion plain manifests with a little template to replace numbers of CPUs or memory is much easier.
I understand that helm is good for, just trying something, but you could also provide a directory that has some yaml files which you can deploy with kubectl create -f dir. Maybe something that can help. What do you think?

[kolbis]

I understand your point. That said, I think it has many benefits which overcome the downfalls such as "hard to read templating engine".
For us, microservice may have dependency in redis, elastic, mssql and etc.
Coordinating such a deployment is easier with helm as you are reusing charts.

I think supporting helm for skipper would be a nice add-on.

[szuecs]

https://github.com/baez90/skipper-helm/ can be a start

[szuecs]

@kolbis I wrote documentation. Would you mind checking if this helps you in order to get faster started?
The files, for example {!kubernetes/deploy/demo/svc.yaml!} will be rendered in the markdown, but not in the diff.

Basically the workflow would be:

git clone https://github.com/zalando/skipper.git
cd skipper
kubectl create -f docs/kubernetes/deploy/deployment

After that you only need to direct somehow LB (ALB, GLB, F5, ...) traffic to Kubernetes worker nodes port 9999 and use health check GET /kube-system/healthz to sort out where skipper is running.

[szuecs]

It's now available at https://opensource.zalando.com/skipper/kubernetes/ingress-controller/#install-skipper-as-ingress-controller

[kolbis]

Thanks!
I will have a look.
If we will decide to create a chart, we will open source it.
I Will update

[szuecs]

If you like create a doc PR, maybe someone find it useful, too

[szuecs]

I had a discussion about helm v3 an @CrankyDragon convinced me that it makes sense to create an official helm chart to simplify testing the ingress stack.
We can have some knobs to change to make sure you can easily test a simple setup and one with all batteries included.

It might not be the “production ready” helm chart, because you would need to understand the workloads better before choosing the right parameters.