Storage provisioner broken for multinode mode

[R-omk]

It seems that the hostpath provisioner is completely unsuitable for multinode installations.

It must create and prepare directories on the exact host on which a pv is requested. In addition, it must add nodeaffinity for every hostpath pv exactly the same as it happens at https://github.com/rancher/local-path-provisioner

Steps to reproduce the issue:

1. minikube create -n 3
2. install any statefulset that requests pv and has pod anti affinity rule

minikube version: v1.22.0
storage-provisioner: image: 'gcr.io/k8s-minikube/storage-provisioner:v5'

[afbjorklund]

It seems that the hostpath provisioner is completely unsuitable for multinode installations.

I think this is a known limitation. Possibly somewhat undocumented, but not surprising.

Theoretically it could work, if "something else" was distributing the files between nodes ?

But node affinity sounds like a good addition. (multi-node was not really part of the spec)

Unfortunately minikube's local storage provisioner is somewhat old and unmaintained 😔

[sharifelgamal]

Yeah, this is a known issue. Thank you for filing this. Help wanted!

[afbjorklund]

afaik, kind just uses the rancher local-path-provisioner (although somewhat customized) :
https://github.com/kubernetes-sigs/kind/blob/main/pkg/build/nodeimage/const_storage.go

• https://github.com/rancher/local-path-provisioner (as used by k3s)

• https://github.com/kubernetes/minikube/tree/master/pkg/storage (since e969ebe v0.17.0)

See #3628 for history

[k8s-triage-robot]

The Kubernetes project currently lacks enough contributors to adequately respond to all issues and PRs.

This bot triages issues and PRs according to the following rules:

• After 90d of inactivity, lifecycle/stale is applied
• After 30d of inactivity since lifecycle/stale was applied, lifecycle/rotten is applied
• After 30d of inactivity since lifecycle/rotten was applied, the issue is closed

You can:

• Mark this issue or PR as fresh with /remove-lifecycle stale
• Mark this issue or PR as rotten with /lifecycle rotten
• Close this issue or PR with /close
• Offer to help out with Issue Triage

Please send feedback to sig-contributor-experience at kubernetes/community.

/lifecycle stale

[R-omk]

/remove-lifecycle stale

[Poweranimal]

Is there an alternative storage driver one can use for minikube that supports multi-node setups?

[R-omk]

install_provisioner:
  curl https://raw.githubusercontent.com/rancher/local-path-provisioner/master/deploy/local-path-storage.yaml | sed 's/\/opt\/local-path-provisioner/\/var\/opt\/local-path-provisioner/ ' | kubectl apply -f -
  kubectl patch storageclass standard -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"false"}}}'
  kubectl patch storageclass local-path -p '{"metadata": {"annotations":{"storageclass.kubernetes.io/is-default-class":"true"}}}'

[Poweranimal]

@R-omk Thanks for your quick reply.

So this installs the rancher local-path-provisioner into my minikube cluster.
Then one sets the local-path-provisioner to the default storage class.

I will install it in the upcoming days in my cluster and will let you know, if I face any troubles ;)

[Poweranimal]

@R-omk Yes, the local-path-provisioner solved all my multi node issues.
Thanks a lot!

[klaases]

Hi @R-omk, we continue to seek help wanted for this issue.

[sharifelgamal]

I've frozen this issue we won't end up accidentally closing it due to age.

[imrj]

will this be fixed at some point? I am using the workaround from @R-omk but was hoping for a formal fix in a future release

[klaases]

Hi @imrj, it looks like @presztak has created a PR for this issue, hopefully this helps, please feel free to review #15062