Minikube: Start --kubernetes-version="v1.3.5" does not work.

[kramarz]

Minikube version: 0.10.0

Environment:

• OS (e.g. from /etc/os-release): Ubuntu 16.04 LTS
• VM Driver: virtualbox

What happened:

$minikube start  --kubernetes-version="v1.3.5"
Starting local Kubernetes cluster...
Kubectl is now configured to use the cluster.
$kubectl cluster-info
The connection to the server 192.168.99.102:8443 was refused - did you specify the right host or port?

What you expected to happen:
As explained in help --kubernetes-version="v1.3.5" should be valid option, as it used to be in 0.9.0
and as it stated in $ minikube help start
--kubernetes-version="v1.3.5+$Format:%h$": The kubernetes version that the minikube VM will (ex: v1.2.3) OR a URI which contains a localkube binary (ex: https://storage.googleapis.com/minikube/k8sReleases/v1.3.0/localkube-linux-amd64)

How to reproduce it (as minimally and precisely as possible):
Execute minikube start --kubernetes-version="v1.3.5"

Anything else do we need to know:
I know it is default, but i want to have it in my setup script so future updates don't change it.
As a workaroud I have copied value from help "v1.3.5+\$Format:%h\$" and it works but I have no idea what it means. Providing documentation in help for this syntax would be enough for fixing this issue.

[brancz]

Getting the same with v1.4.0-beta.2 also minikube version 0.10.0 on OS X with xhyve. Reverted back to 0.9.0 and it works.

[webwurst]

When using minikube start --vm-driver kvm --memory 2048 --kubernetes-version="v1.3.5" with minikube v0.10.0 the following error shows up:

$ minikube logs
==> /var/lib/localkube/localkube.err <==

==> /var/lib/localkube/localkube.out <==
unknown flag: --node-ip
Usage of /usr/local/bin/localkube:
      --alsologtostderr[=false]: log to standard error as well as files
      --apiserver-address=0.0.0.0: The address the apiserver will listen securely on
      --apiserver-insecure-address=127.0.0.1: The address the apiserver will listen insecurely on
[...]

update:
error for v1.3.6:

$ minikube start --vm-driver kvm --memory 2048 --kubernetes-version="v1.3.6"                                        591ms 
Starting local Kubernetes cluster...
Kubectl is now configured to use the cluster.
 D/g/g/kubernetes-registry/ingress-controller $ minikube logs                                                                                                 45s 307ms 
==> /var/lib/localkube/localkube.err <==

==> /var/lib/localkube/localkube.out <==

 D/g/g/kubernetes-registry/ingress-controller $ kubectl cluster-info                                                                                              265ms 
The connection to the server 192.168.42.87:8443 was refused - did you specify the right host or port?

error for v1.3.7:

$ minikube start --vm-driver kvm --memory 2048 --kubernetes-version="v1.3.7"                                        418ms 
Starting local Kubernetes cluster...
Kubectl is now configured to use the cluster.
 D/g/g/kubernetes-registry/ingress-controller $ minikube logs                                                                                                 42s 847ms 
==> /var/lib/localkube/localkube.err <==

==> /var/lib/localkube/localkube.out <==
/usr/local/bin/localkube: line 1: syntax error: unexpected redirection

[webwurst]

ref #532

[r2d4]

Thanks for the detailed report. I've opened up a PR to fix this issue. We were always passing --node-ip to the kubelet, which previous versions of kubernetes don't know about so it was breaking. The proposed fix is to only pass it in if specified. This will fix the 1.3.5, 1.3.6, and beta releases. We haven't released a 1.3.7 localkube version so that still wont work :)

We can do a 0.10.1 release later today to address this. Sorry about the inconvenience!

[jimmidyson]

Technically it's a problem with localkube, not kubernetes. The node IP being valid is used by a number of tools so I've proposed that we pass the node IP if the flag is valid for that version of kubernetes.

[r2d4]

Sorry for the delay in fixing, its been a little trickier than I thought. So I've fixed the 1.3.x releases, however theres still a bug I'm working on with the 1.4.0-betas. I'm also working on a patch for 0.10.1 release, but in the meantime, you can get these immediately by doing a rm -r ~/.minikube/cache/localkube and then the fixed versions are as follows

minikube start --kubernetes-version=v1.3.3
minikube start --kubernetes-version=v1.3.4
minikube start --kubernetes-version=v1.3.5
minikube start --kubernetes-version=v1.3.6
minikube start --kubernetes-version=v1.3.7

In addition, I've set up jenkins jobs to run regression tests for each release so we don't inadvertently break old localkube versions

[r2d4]

Fixed the beta issue. You should now be able to run

minikube start --kubernetes-version=v1.4.0-beta.8

These will be documented in the patch

[Starefossen]

Looking forward to start playing around with the latest v1.4.0 beta on Minikube 0.10 😄

[dlorenc]

I think this is fixed now.