specifying node-cidr-mask-size as a controllerManagerExtra arg did not take precedence

[tmjd]

What keywords did you search in kubeadm issues before filing this one?

controllerManagerExtraArgs, node-cidr-mask-size

Is this a BUG REPORT or FEATURE REQUEST?

BUG REPORT

Versions

kubeadm version (use kubeadm version): v1.9.3

Environment:

• Kubernetes version (use kubectl version): v1.9.3
• Cloud provider or hardware configuration: VirtualBox
• OS (e.g. from /etc/os-release):

  NAME="Container Linux by CoreOS"
  ID=coreos
  VERSION=1576.4.0
  VERSION_ID=1576.4.0
  BUILD_ID=2017-12-06-0449
  PRETTY_NAME="Container Linux by CoreOS 1576.4.0 (Ladybug)"
  ANSI_COLOR="38;5;75"
  HOME_URL="https://coreos.com/"
  BUG_REPORT_URL="https://issues.coreos.com"
  COREOS_BOARD="amd64-usr"
  

• Kernel (e.g. uname -a): Linux k8s-master 4.13.16-coreos-r2 kubeadm join on slave node fails preflight checks #1 SMP Wed Dec 6 04:27:34 UTC 2017 x86_64 Intel(R) Xeon(R) CPU E3-1505M v5 @ 2.80GHz GenuineIntel GNU/Linux
• Others:

What happened?

I was specifying node-cidr-mask-size in a kubeadm config file under controllerManagerExtraArgs to set the flag. kube-controller-manager was started with 2 instances of --node-cidr-mask-size being passed to it. The 2nd instance was the default that kubeadm adds and overrode the value I was attempting to set.

What you expected to happen?

kube-controller-manager should have been started with one instance of --node-cidr-mask-size or at least the one I specified would be the 2nd so that it would take precedence.

How to reproduce it (as minimally and precisely as possible)?

1. Run kubeadm init --config=myconfig.yaml with the following as the contents of the file myconfig.yaml

apiVersion: kubeadm.k8s.io/v1alpha1
kind: MasterConfiguration
api:  
  advertiseAddress: "fd10::101"
networking:
  serviceSubnet: fd30::0/110   
  podSubnet: "fd20:0::/120"    
controllerManagerExtraArgs:
  node-cidr-mask-size: "120" 

2. Check the flags that kube-controller-manager is started with.

Anything else we need to know?

This was an issue for me when trying to use kubeadm to setup an IPv6 cluster and specify a podSubnet.

[XANi]

still present in 1.9.6.

It should really be just an option under networking key, currently it just puts --node-cidr-mask-size in config twice, at start and at end

[jstangroome]

@tmjd I've worked around it by moving the networking.podSubnet to controllerManagerExtraArgs.cluster-cidr and adding allocate-node-cidrs too, just like the kubeadm code would have if podSubnet had been specified directly.

Modifying your original example, it becomes:

apiVersion: kubeadm.k8s.io/v1alpha1
kind: MasterConfiguration
api:  
  advertiseAddress: "fd10::101"
networking:
  serviceSubnet: fd30::0/110   
  # MOVED BELOW podSubnet: "fd20:0::/120"    
controllerManagerExtraArgs:
  allocate-node-cidrs: "true"
  cluster-cidr: "fd20:0::/120"
  node-cidr-mask-size: "120"

This workaround is essentially circumventing kubeadm's if statement here:
https://github.com/kubernetes/kubernetes/blob/86a58202b68d04b2e31b56db80b4d2a4dec77c93/cmd/kubeadm/app/phases/controlplane/manifests.go#L336-L342

I doesn't appear that kubeadm uses the podSubnet for anything else other than validation but I could be wrong.

Would still prefer to be able to override the node cidr mask size in a more supported fashion though.

[jstangroome]

@tmjd Fix should land in next kubeadm release from master.

[Hazmes]

What is the status on this?
I am currently setting up a Kubernetes Cluster with:

kubeadm version: &version.Info{Major:"1", Minor:"10", GitVersion:"v1.10.4", GitCommit:"5ca598b4ba5abb89bb773071ce452e33fb66339d", GitTreeState:"clean", BuildDate:"2018-06-06T08:00:59Z", GoVersion:"go1.9.3", Compiler:"gc", Platform:"linux/amd64"}

The Issue is still occurring here.