search

HCL-TECH-SOFTWARE / connections-automation

Deployment and upgrade automation scripts for HCL Connections 7.0 based on Ansible
Apache License 2.0
17 stars 31 forks source link

Kubernetes 1.24: kubeadmin config wrong api version #221

Closed stoeps13 closed 1 year ago

stoeps13 commented 1 year ago

kubeadmin 1.24+ uses apiVersion: kubeadm.k8s.io/v1beta3 instead of apiVersion: kubeadm.k8s.io/v1beta2

So the kubeadmin init fails

sabrina-yee commented 1 year ago

@stoeps13 what is the Kubernetes version set in your all.yml for kubernetes_version? We use 1.24.1 as the default so want to use what you set to attempt to reproduce the issue. Thanks.

stoeps13 commented 1 year ago

@sabrina-yee I haven't set a version, it uses the default 1.24.1

I tried manually with the generated config.yaml it shows wrong api version.

sabrina-yee commented 1 year ago

Thanks for the info, need to investigate further as we use the default internally as well but haven't run into this issue.

sabrina-yee commented 1 year ago

@stoeps13 we ran into this issue today along with issue #223. The problem in our case is that containerd was installed as part of docker (via the setup-docker-registry.yml playbook) rather than using setup-containerd.yml from setup-component-pack-complete-harbor.yml. The docker one creates a different /etc/containerd/config.toml with this line in it:

disabled_plugins = ["cri"]

that leads to [ERROR CRI]: container runtime is not running. Commenting out this line will address both this issue and #223.

stoeps13 commented 1 year ago

Hi Sabrina, I found the issue too. There is a new playbook named https://github.com/HCL-TECH-SOFTWARE/connections-automation/blob/main/playbooks/setup-component-pack-complete-harbor.yml, I did the first try with the old playbook which fails with the described errors. Regards Christoph