Configure Flannel networking tasks fails on Ubuntu 20.04 about ansible-role-kubernetes HOT 6 CLOSED

I have verified that all nodes have the same certificate-authority-data in their /root/.kube/config.

from ansible-role-kubernetes.

A shot in the dark, but I added a full package update and reboot to see if that solved the issue. It was unsuccessful.

---
- hosts: kube
  become: true
  handlers:
    - name: reboot
      reboot:

  pre_tasks:
    # adding to see if updating all packages will resolve the issue of 
    # Configure Flannel networking task failing on worker nodes.
    - name: update all packages # noqa 403
      apt:
        name: '*'
        state: latest
        update_cache: true
      notify: reboot

    # ensure handlers are flushed before moving on to geerlingguys roles. 
    - name: flush handlers
      meta: flush_handlers

  # Geerlingguy's roles per Ansible for Kubernetes page 77 (2021Sep30).
  roles:
    - geerlingguy.security
    - geerlingguy.docker
    - geerlingguy.swap
    - geerlingguy.kubernetes

from ansible-role-kubernetes.

I'm thinking maybe this issue should be in https://github.com/geerlingguy/ansible-for-kubernetes?

from ansible-role-kubernetes.

I ran an ansible ad-hoc command to get all of the /etc/kubernetes/admin.conf files from my nodes so I could inspect them all on my local machine:

ansible -m fetch -a "src=/etc/kubernetes/admin.conf dest=/tmp/fetch" -i inventory/hosts.yml all -b

I did a diff across all the files and saw that I was mistaken. The certificate-authority-data was different across the board. As a quick test I copied the certificate-authority-data from my master node's admin.conf on my local to my second node's admin.conf on my local and then pushed that file back out to node02. I sshed to node02 and switched to the root user. Then just a kubectl get nodes and boom, no certificate errors. However, I did get an error about not being a logged in user.

root@node05:~# kubectl get nodes
error: You must be logged in to the server (Unauthorized)

from ansible-role-kubernetes.

Replaced all worker nodes with the exact same /etc/kubernetes/admin.conf I fetched to my local from node01 one with a quick script:

#!/bin/bash
for i in 2 3 4 5
do
ansible -m copy -a "src=/tmp/fetch/node01/etc/kubernetest/admin.conf dest=/etc/kubernetes/admin.conf" -i inventory/hosts.yml all -b --limit node0$1
done

Then ran the playbook again and was met with a completed execution but still only seeing node01 when I check on all the nodes:

root@node01:~# kubectl get nodes
NAME                        STATUS   ROLES    AGE   VERSION
node01.test.local       Ready      <none>    45h     v1.20.11

from ansible-role-kubernetes.

I completely missed the kubernetes_role in the inventory on page 74 of Ansible for Kubernetes. I delete and redeployed my nodes in my test environment. I modified my inventory like so:

all:
  children:
    kube:
      children:
        kubemaster:
        kubeworker:
    kubemaster:
      hosts:
        node01:
    kubeworker:
      hosts:
        node0[2:5]:

inventory/group_vars/kubemaster.yml

---
# Kubernetes master configuration.
kubernetes_role: master

inventory/group_vars/kubeworker.yml

---
# Kubernetes worker configuration.
kubernetes_role: node

I reran the playbook and logged back in node one and I could see all of the worker nodes in the cluster! 🎉

root@node01:~# kubectl get nodes
NAME                    STATUS   ROLES                       AGE   VERSION
node01.test.local   Ready      control-plane,master       60s   v1.20.11
node02.test.local   Ready      <none>                       32s   v1.20.11
node03.test.local   Ready      <none>                       33s   v1.20.11
node04.test.local   Ready      <none>                       33s   v1.20.11
node05.test.local   Ready      <none>                       31s   v1.20.11

Sorry for the confusion and opening up a ticket unnecessarily. Thanks for all the work you do!

from ansible-role-kubernetes.