Create cluster is not working with the latest k8s version

[Zebradil]

I think it is somehow related to #162.

I tried to switch to the latest available k8s version, but the cluster creation process failed with the following info:

2018/10/07 23:17:10 Creating new cluster

NAME:demo
MASTERS: 1
WORKERS: 1
ETCD NODES: 0
HA: false
ISOLATED ETCD: false
2018/10/07 23:17:11 creating server 'demo-master-01'...
  --- [======================================] 100%
2018/10/07 23:17:33 Created node 'demo-master-01' with IP 78.46.183.151
2018/10/07 23:17:33 creating server 'demo-worker-01'...
  --- [======================================] 100%
2018/10/07 23:18:04 Created node 'demo-worker-01' with IP 78.46.194.165
2018/10/07 23:18:04 sleep for 10s...
demo-master-01       : wireguard configured               39.1% [====>---------]
demo-worker-01       : wireguard configured               64.3% [========>-----]
2018/10/07 23:21:37 run failed
command:kubeadm token create --print-join-command
stdout:
stderr:Process exited with status 1

I tried to figure out what's wrong but got stuck. It seems like --debug flag changes nothing.
I checked created nodes and didn't even find /root/master-config.yaml file which (if I'm not mistaken) should be already there before kubeadm token create --print-join-command command.

I understand, that there is only one supported version at the moment, but maybe someone can direct me where to dig and I can resolve the issue.

[Zebradil]

The version which failed is 0.12*. I tried 0.10* and it works (I needed it, because it supports ConfigMap with binaryData field).

[JohnnyQQQQ]

Can you give me the specific command you typed in to start the installation?

[Zebradil]

I pulled the last version of hetzner-kube from master, changed k8s-version:

diff --git a/pkg/clustermanager/provision_node.go b/pkg/clustermanager/provision_node.go
index 2cd4399..1fa6a58 100644
--- a/pkg/clustermanager/provision_node.go
+++ b/pkg/clustermanager/provision_node.go
@@ -10,7 +10,7 @@ import (
 const maxErrors = 3

 // K8sVersion is the version that will be used to install kubernetes
-var K8sVersion = flag.String("k8s-version", "1.9.6-00",
+var K8sVersion = flag.String("k8s-version", "1.12*",
        "The version of the k8s debian packages that will be used during provisioning")

 // NodeProvisioner provisions all basic packages to install docker, kubernetes and wireguard

and created cluster with hetzner-kube cluster create --debug --name demo --ssh-key demo

[rvasilev]

@Zebradil configuration files of k8 1.9 and 1.12 are incompatible, that's not enough just to change the version number. But you could install 1.9 and gradually upgrade to 1.12.

https://kubernetes.io/docs/reference/setup-tools/kubeadm/kubeadm-upgrade/

Config changes/deprecations:
1.9-1.11: https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG-1.11.md#new-deprecations
1.12: https://github.com/kubernetes/kubernetes/blob/master/CHANGELOG-1.12.md#action-required-4

[Zebradil]

Thanks! I'll try it.

[gridworkz]

So what should we do here now? I'm getting the same: stderr:Process exited with status 1

[mavimo]

@gridworkz there are a PoC that run the new cluster on this PR #208 but need some more work, we'll update the project state when the new cluster config will work as expected ;)