C4W3 Lab1 Error in creating kubeflow

Course Q&A Machine Learning Engineering for Production(MLOps) Deploying Machine Learning Models in Production
1

When I create kubeflow on minikube, I got the following output:

(base) ~/ml/kfp $ kubectl apply -k “github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=$PIPELINE_VERSION”

namespace/kubeflow unchanged

customresourcedefinition.apiextensions.k8s.io/clusterworkflowtemplates.argoproj.io unchanged

customresourcedefinition.apiextensions.k8s.io/cronworkflows.argoproj.io unchanged

customresourcedefinition.apiextensions.k8s.io/workfloweventbindings.argoproj.io unchanged

customresourcedefinition.apiextensions.k8s.io/workflows.argoproj.io unchanged

customresourcedefinition.apiextensions.k8s.io/workflowtemplates.argoproj.io unchanged

serviceaccount/kubeflow-pipelines-cache-deployer-sa unchanged

clusterrole.rbac.authorization.k8s.io/kubeflow-pipelines-cache-deployer-clusterrole unchanged

clusterrolebinding.rbac.authorization.k8s.io/kubeflow-pipelines-cache-deployer-clusterrolebinding unchanged

unable to recognize “github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=1.7.0”: no matches for kind “CustomResourceDefinition” in version “apiextensions.k8s.io/v1beta1”

unable to recognize “github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=1.7.0”: no matches for kind “CustomResourceDefinition” in version “apiextensions.k8s.io/v1beta1”

unable to recognize “github.com/kubeflow/pipelines/manifests/kustomize/cluster-scoped-resources?ref=1.7.0”: no matches for kind “CustomResourceDefinition” in version “apiextensions.k8s.io/v1beta1”

Please help.

2

Instruction at notebook says use kind. Local cluster with minikube did not work for kubeflow with its minimum resource needs.

minikube start --cpus 4 --memory 8096 --disk-size=40g

Got this message at terminal:

Your cgroup does not allow setting memory.
:black_small_square: More information: Post-installation steps for Linux | Docker Documentation

:x: Exiting due to RSRC_OVER_ALLOC_MEM: Requested memory allocation 8096MB is more than your system limit 6190MB.

Changing the setting seems complicated.

Installed kind. But after creating cluster with kind, machine becomes excruciatingly slow. Probably it is taking away lot of cpu and memory. I am using HP i5 with 4 cores (8 logical) and 8 GB RAM. Clearly under-powered for kubeflow.

Question to community: Is there another local cluster provider like minikube and kind that can run kubeflow in my machine? Or is it then an excuse for me to upgrade laptop?

https://v0-2.kubeflow.org/docs/started/getting-started-minikube/

Regards.
Sanjoy

3

Alternatively can colab pro terminal be used to install and run kubeflow with one of the local k8 cluster provider like minikube or kind?

Please share your experience if you have used it.

4

I have the same problem running the kubeflow lab with kind

Creating the kind cluster is not a problem

type or paste code here% kubectl get nodes -o wide
NAME                 STATUS   ROLES                  AGE   VERSION   INTERNAL-IP   EXTERNAL-IP   OS-IMAGE       KERNEL-VERSION      CONTAINER-RUNTIME
kind-control-plane   Ready    control-plane,master   13m   v1.23.4   172.20.0.2    <none>        Ubuntu 21.10   5.4.0-100-generic   containerd://1.5.10

However applying the manifest fails. It seems, that this is a known issue with no solution: no matches for kind "CustomResourceDefinition" in version "apiextensions.k8s.io/v1beata1" · Issue #2028 · kubeflow/manifests · GitHub

How can I fix this?

//edit: found a way:

You have to create a kind cluster like this:

kind create cluster --image kindest/node:v1.21.2

Then everything works out as described in the notebook

4 Likes