Inventory file (inventory/k8s.yaml):
plugin: kubernetes.core.k8s
connections:
- kubeconfig: ~/.kube/config
context: 'cluster-2'
Task file (roles/common/tasks/main.yaml):
# Method 1: Using `kubernetes.core` plugin to list the pod names:
- name: Get a list of all pods from any namespace
kubernetes.core.k8s_info:
kind: Pod
register: pod_list
- name: Print pod names
debug:
msg: "pod_list: {{ pod_list | json_query('resources[*].metadata.name') }} "
# Method 2: Using `shell` command to list the pod names:
- name: Get node names
shell: kubectl get pods
register: pod_list2
- name: Print pod names
debug:
msg: "{{ pod_list2.stdout }}"
Ansible config (ansible.cfg):
[inventory]
enable_plugins = host_list, auto, yaml, ini, kubernetes.core.k8s
Main file (main.yaml):
---
- hosts: localhost
gather_facts: false
collections:
- azure.azcollection
- kubernetes.core
roles:
- "common"
Running command to execute task: ansible-playbook main.yaml -i cluster-2/k8s.yaml -e role=common -e cluster_name=cluster-2
Question: I am running the above configs to run get the pods from the remote cluster mentioned in the inventory file. But, the problem is, I am still getting the pod names from the local cluster and not the cluster-2 in Method 1 and 2.
k8s plugin should get the list of pods from cluster-2 as described in the inventory file. How can I connect to remote kubernetes cluster?
I also checked output with -vvvv:
ansible-playbook [core 2.14.0]
config file = /Users/test/u/apps/ansible.cfg
configured module search path = ['/Users/test/.ansible/plugins/modules', '/usr/share/ansible/plugins/modules']
ansible python module location = /Users/test/Library/Python/3.9/lib/python/site-packages/ansible
ansible collection location = /Users/test/.ansible/collections:/usr/share/ansible/collections
executable location = /Users/test/Library/Python/3.9/bin/ansible-playbook
python version = 3.9.12 (main, Mar 26 2022, 15:52:10) [Clang 13.0.0 (clang-1300.0.29.30)] (/usr/local/opt/[email protected]/bin/python3.9)
jinja version = 3.1.2
libyaml = True
Using /Users/test/u/apps/ansible.cfg as config file
setting up inventory plugins
Loading collection kubernetes.core from /Users/test/.ansible/collections/ansible_collections/kubernetes/core
You're trying to use both the kubernetes inventory plugin and the k8s_info module, and because of that you're getting conflicting results. The two don't have anything to do with each other.
The kubernetes inventory module is -- I think -- a weird beast; it produces an ansible inventory in which the pods in your cluster are presented as Ansible hosts. To see a list of all the pod names in your cluster, you could write a playbook like this:
- hosts: all
gather_facts: false
tasks:
- name: Print pod names
debug:
msg: "{{ inventory_hostname }}"
This will respect the context you've configured in your kubernetes inventory plugin configuration. For example, if I have in inventory/k8s.yaml the following:
plugin: kubernetes.core.k8s
connections:
- kubeconfig: ./kubeconfig
context: 'kind-cluster2'
Then the above playbook will list the pod names from kind-cluster2, regardless of the current-context setting in my kubeconfig file. In my test environment, this produces:
PLAY [all] *********************************************************************
TASK [Print pod names] *********************************************************
ok: [kubernetes] => {
"msg": "kubernetes"
}
ok: [coredns-565d847f94-2shl6_coredns] => {
"msg": "coredns-565d847f94-2shl6_coredns"
}
ok: [coredns-565d847f94-md57c_coredns] => {
"msg": "coredns-565d847f94-md57c_coredns"
}
ok: [kube-dns] => {
"msg": "kube-dns"
}
ok: [etcd-cluster2-control-plane_etcd] => {
"msg": "etcd-cluster2-control-plane_etcd"
}
ok: [kube-apiserver-cluster2-control-plane_kube-apiserver] => {
"msg": "kube-apiserver-cluster2-control-plane_kube-apiserver"
}
ok: [kube-controller-manager-cluster2-control-plane_kube-controller-manager] => {
"msg": "kube-controller-manager-cluster2-control-plane_kube-controller-manager"
}
ok: [kube-scheduler-cluster2-control-plane_kube-scheduler] => {
"msg": "kube-scheduler-cluster2-control-plane_kube-scheduler"
}
ok: [kindnet-nc27b_kindnet-cni] => {
"msg": "kindnet-nc27b_kindnet-cni"
}
ok: [kube-proxy-9chgt_kube-proxy] => {
"msg": "kube-proxy-9chgt_kube-proxy"
}
ok: [local-path-provisioner-684f458cdd-925v5_local-path-provisioner] => {
"msg": "local-path-provisioner-684f458cdd-925v5_local-path-provisioner"
}
PLAY RECAP *********************************************************************
coredns-565d847f94-2shl6_coredns : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
coredns-565d847f94-md57c_coredns : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
etcd-cluster2-control-plane_etcd : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
kindnet-nc27b_kindnet-cni : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
kube-apiserver-cluster2-control-plane_kube-apiserver : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
kube-controller-manager-cluster2-control-plane_kube-controller-manager : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
kube-dns : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
kube-proxy-9chgt_kube-proxy : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
kube-scheduler-cluster2-control-plane_kube-scheduler : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
kubernetes : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
local-path-provisioner-684f458cdd-925v5_local-path-provisioner : ok=1 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
The key point here is that your inventory will consist of a list of pods. I've never found this particularly useful.
k8s_info moduleThe k8s_info queries a kubernetes cluster for a list of objects. It doesn't care about your inventory configuration -- it will run on whichever target host you've defined for your play (probably localhost) and perform the rough equivalent of kubectl get <whatever>. If you want to use an explicit context, you need to set that as part of your module parameters. For example, to see a list of pods in kind-cluster2, I could use the following playbook:
- hosts: localhost
gather_facts: false
tasks:
- kubernetes.core.k8s_info:
kind: pod
kubeconfig: ./kubeconfig
context: kind-cluster2
register: pods
- debug:
msg: "{{ pods.resources | json_query('[].metadata.name') }}"
Which in my test environment produces as output:
PLAY [localhost] ***************************************************************
TASK [kubernetes.core.k8s_info] ************************************************
ok: [localhost]
TASK [debug] *******************************************************************
ok: [localhost] => {
"msg": [
"coredns-565d847f94-2shl6",
"coredns-565d847f94-md57c",
"etcd-cluster2-control-plane",
"kindnet-nc27b",
"kube-apiserver-cluster2-control-plane",
"kube-controller-manager-cluster2-control-plane",
"kube-proxy-9chgt",
"kube-scheduler-cluster2-control-plane",
"local-path-provisioner-684f458cdd-925v5"
]
}
PLAY RECAP *********************************************************************
localhost : ok=2 changed=0 unreachable=0 failed=0 skipped=0 rescued=0 ignored=0
In conclusion: you probably want to use k8s_info rather than the inventory plugin, and you'll need to configure the module properly by setting the context (and possibly the kubeconfig) parameters when you call the module.
Is there any way I can define context and kubeconfig outside of the tasks (globally) if I am using k8s_info module?
According to the documentation, you could set the K8S_AUTH_KUBECONFIG and K8S_AUTH_CONTEXT environment variables if you want to globally configure the settings for the k8s_info module. You could also write your task like this:
- kubernetes.core.k8s_info:
kind: pod
kubeconfig: "{{ k8s_kubeconfig }}"
context: "{{ k8s_context }}"
register: pods
And then define the k8s_kubeconfig and k8s_context variables somewhere else in your Ansible configuration (e.g., as group vars). This makes it easy to retarget things to a different cluster with only a single change.