Merge pull request #11793 from VannTen/cleanup/ci_testcases_better_callback

Use debug stdout callback in ci rather than manual debug

.gitlab-ci.yml

@@ -10,6 +10,7 @@ variables:
   FAILFASTCI_NAMESPACE: 'kargo-ci'
   GITLAB_REPOSITORY: 'kargo-ci/kubernetes-sigs-kubespray'
   ANSIBLE_FORCE_COLOR: "true"
+  ANSIBLE_STDOUT_CALLBACK: "debug"
   MAGIC: "ci check this"
   GS_ACCESS_KEY_ID: $GS_KEY
   GS_SECRET_ACCESS_KEY: $GS_SECRET

tests/testcases/010_check-apiserver.yml

@@ -13,9 +13,6 @@
     delay: 5
     until: apiserver_response is success
 
-  - debug:  # noqa name[missing]
-      msg: "{{ apiserver_response.json }}"
-
   - name: Check API servers version
     assert:
       that:

tests/testcases/015_check-nodes-ready.yml

@@ -1,18 +1,10 @@
 ---
 - name: Testcases checking nodes
   hosts: kube_control_plane[0]
+  vars:
+    bin_dir: /usr/local/bin
   tasks:
 
-  - name: Force binaries directory for Flatcar Container Linux by Kinvolk
-    set_fact:
-      bin_dir: "/opt/bin"
-    when: ansible_os_family in ["Flatcar", "Flatcar Container Linux by Kinvolk"]
-
-  - name: Force binaries directory for other hosts
-    set_fact:
-      bin_dir: "/usr/local/bin"
-    when: not ansible_os_family in ["Flatcar", "Flatcar Container Linux by Kinvolk"]
-
   - import_role:  # noqa name[missing]
       name: cluster-dump
 
@@ -21,9 +13,6 @@
     changed_when: false
     register: get_nodes
 
-  - debug:  # noqa name[missing]
-      msg: "{{ get_nodes.stdout.split('\n') }}"
-
   - name: Check that all nodes are running and ready
     command: "{{ bin_dir }}/kubectl get nodes --no-headers -o yaml"
     changed_when: false

tests/testcases/020_check-pods-running.yml

@@ -1,28 +1,16 @@
 ---
 - name: Testcases checking pods
   hosts: kube_control_plane[0]
+  vars:
+    bin_dir: /usr/local/bin
   tasks:
 
-  - name: Force binaries directory for Flatcar Container Linux by Kinvolk
-    set_fact:
-      bin_dir: "/opt/bin"
-    when: ansible_os_family in ["Flatcar", "Flatcar Container Linux by Kinvolk"]
-
-  - name: Force binaries directory for other hosts
-    set_fact:
-      bin_dir: "/usr/local/bin"
-    when: not ansible_os_family in ["Flatcar", "Flatcar Container Linux by Kinvolk"]
-
   - import_role:  # noqa name[missing]
       name: cluster-dump
 
   - name: Check kubectl output
     command: "{{ bin_dir }}/kubectl get pods --all-namespaces -owide"
     changed_when: false
-    register: get_pods
-
-  - debug:  # noqa name[missing]
-      msg: "{{ get_pods.stdout.split('\n') }}"
 
   - name: Check that all pods are running and ready
     command: "{{ bin_dir }}/kubectl get pods --all-namespaces --no-headers -o yaml"
@@ -35,13 +23,7 @@
     - '(run_pods_log.stdout | from_yaml)["items"] | map(attribute = "status.containerStatuses") | map("map", attribute = "ready") | map("min") | min'
     retries: 30
     delay: 10
-    failed_when: false
 
   - name: Check kubectl output
     command: "{{ bin_dir }}/kubectl get pods --all-namespaces -owide"
     changed_when: false
-    register: get_pods
-
-  - debug:  # noqa name[missing]
-      msg: "{{ get_pods.stdout.split('\n') }}"
-    failed_when: not run_pods_log is success

tests/testcases/030_check-network.yml

@@ -4,17 +4,9 @@
   vars:
     test_image_repo: registry.k8s.io/e2e-test-images/agnhost
     test_image_tag: "2.40"
+    bin_dir: "/usr/local/bin"
 
   tasks:
-  - name: Force binaries directory for Flatcar Container Linux by Kinvolk
-    set_fact:
-      bin_dir: "/opt/bin"
-    when: ansible_os_family in ["Flatcar", "Flatcar Container Linux by Kinvolk"]
-
-  - name: Force binaries directory for other hosts
-    set_fact:
-      bin_dir: "/usr/local/bin"
-    when: not ansible_os_family in ["Flatcar", "Flatcar Container Linux by Kinvolk"]
 
   - name: Check kubelet serving certificates approved with kubelet_csr_approver
     when:
@@ -29,9 +21,6 @@
       register: csr_json
       changed_when: false
 
-    - debug:  # noqa name[missing]
-        var: csrs
-
     - name: Check there are csrs
       assert:
         that: csrs | length > 0
@@ -67,17 +56,13 @@
       when: get_csr.stdout_lines | length > 0
       changed_when: certificate_approve.stdout
 
-    - debug:  # noqa name[missing]
-        msg: "{{ certificate_approve.stdout.split('\n') }}"
-
-
   - name: Create test namespace
     command: "{{ bin_dir }}/kubectl create namespace test"
     changed_when: false
 
   - name: Run 2 agnhost pods in test ns
     command:
-      cmd: "{{ bin_dir }}/kubectl apply -f -"
+      cmd: "{{ bin_dir }}/kubectl apply --namespace test -f -"
       stdin: |
         apiVersion: apps/v1
         kind: Deployment
@@ -107,52 +92,35 @@
                     type: RuntimeDefault
     changed_when: false
 
-  - import_role:  # noqa name[missing]
-      name: cluster-dump
-
   - name: Check that all pods are running and ready
+    vars:
+      pods: "{{ (pods_json.stdout | from_json)['items'] }}"
     block:
     - name: Check Deployment is ready
-      command: "{{ bin_dir }}/kubectl rollout status deploy --namespace test agnhost --timeout=180"
+      command: "{{ bin_dir }}/kubectl rollout status deploy --namespace test agnhost --timeout=180s"
       changed_when: false
-    rescue:
     - name: Get pod names
       command: "{{ bin_dir }}/kubectl get pods -n test -o json"
       changed_when: false
-      register: pods
+      register: pods_json
 
-  - name: Get running pods
-    command: "{{ bin_dir }}/kubectl get pods -n test -o
-            jsonpath='{range .items[?(.status.phase==\"Running\")]}{.metadata.name} {.status.podIP} {.status.containerStatuses} {end}'"
-    changed_when: false
-    register: running_pods
+    - name: Check pods IP are in correct network
+      assert:
+        that: pods
+              | selectattr('status.phase', '==', 'Running')
+              | selectattr('status.podIP', 'ansible.utils.in_network', kube_pods_subnet)
+              | length == 2
+
+    - name: Curl between pods is working
+      command: "{{ bin_dir }}/kubectl -n test exec {{ item[0].metadata.name }} -- curl {{ item[1].status.podIP }}:8080"
+      with_nested:
+      - "{{ pods }}"
+      - "{{ pods }}"
+    rescue:
+    - name: List pods cluster-wide
+      command: "{{ bin_dir }}/kubectl get pods --all-namespaces -owide"
+      changed_when: false
 
-  - name: Check kubectl output
-    command: "{{ bin_dir }}/kubectl get pods --all-namespaces -owide"
-    changed_when: false
-    register: get_pods
-
-  - debug:  # noqa name[missing]
-      msg: "{{ get_pods.stdout.split('\n') }}"
-
-  - name: Set networking facts
-    set_fact:
-      kube_pods_subnet: 10.233.64.0/18
-      pod_names: "{{ (pods.stdout | from_json)['items'] | map(attribute='metadata.name') | list }}"
-      pod_ips: "{{ (pods.stdout | from_json)['items'] | selectattr('status.podIP', 'defined') | map(attribute='status.podIP') | list }}"
-      pods_running: |
-        {% set list = running_pods.stdout.split(" ") %}
-        {{ list }}
-
-  - name: Check pods IP are in correct network
-    assert:
-      that: item | ansible.utils.ipaddr(kube_pods_subnet)
-    when:
-    - item in pods_running
-    with_items: "{{ pod_ips }}"
-
-  - name: Curl between pods is working
-    command: "{{ bin_dir }}/kubectl -n test exec {{ item[0] }} -- curl {{ item[1] }}:8080"
-    with_nested:
-    - "{{ pod_names }}"
-    - "{{ pod_ips }}"
+    - import_role:  # noqa name[missing]
+        name: cluster-dump
+    - fail: # noqa name[missing]

tests/testcases/040_check-network-adv.yml

@@ -19,18 +19,9 @@
     agent_report_interval: 10
     netcheck_namespace: default
     netchecker_port: 31081
+    bin_dir: "/usr/local/bin"
 
   tasks:
-    - name: Force binaries directory for Container Linux by CoreOS and Flatcar
-      set_fact:
-        bin_dir: "/opt/bin"
-      when: ansible_os_family in ["Flatcar", "Flatcar Container Linux by Kinvolk"]
-
-    - name: Force binaries directory on other hosts
-      set_fact:
-        bin_dir: "/usr/local/bin"
-      when: not ansible_os_family in ["Flatcar", "Flatcar Container Linux by Kinvolk"]
-
     - import_role:  # noqa name[missing]
         name: cluster-dump
 
@@ -54,97 +45,57 @@
         - netchecker-agent-hostnet
       when: not pods_json is success
 
-    - debug:  # noqa name[missing]
-        var: nca_pod.stdout_lines
-      when: inventory_hostname == groups['kube_control_plane'][0]
-
-    - name: Get netchecker agents
-      uri:
-        url: "http://{{ ansible_default_ipv4.address }}:{{ netchecker_port }}/api/v1/agents/"
-        return_content: true
-      run_once: true
-      delegate_to: "{{ groups['kube_control_plane'][0] }}"
-      register: agents
-      retries: 18
-      delay: "{{ agent_report_interval }}"
-      until: agents.content | length > 0 and
-        agents.content[0] == '{' and
-        agents.content | from_json | length >= groups['k8s_cluster'] | intersect(ansible_play_hosts) | length * 2
-      failed_when: false
-
-    - name: Check netchecker status
-      uri:
-        url: "http://{{ ansible_default_ipv4.address }}:{{ netchecker_port }}/api/v1/connectivity_check"
-        status_code: 200
-        return_content: true
-      delegate_to: "{{ groups['kube_control_plane'][0] }}"
-      run_once: true
-      register: connectivity_check
-      retries: 3
-      delay: "{{ agent_report_interval }}"
-      until: connectivity_check.content | length > 0 and
-        connectivity_check.content[0] == '{'
-      failed_when: false
-      when:
-        - agents.content != '{}'
-
-    - debug:  # noqa name[missing]
-        var: pods_json
-      run_once: true
-
-    - name: Get kube-proxy logs
-      command: "{{ bin_dir }}/kubectl -n kube-system logs -l k8s-app=kube-proxy"
-      when:
-        - inventory_hostname == groups['kube_control_plane'][0]
-        - not connectivity_check is success
-
-    - name: Get logs from other apps
-      command: "{{ bin_dir }}/kubectl -n kube-system logs -l k8s-app={{ item }} --all-containers"
-      when:
-        - inventory_hostname == groups['kube_control_plane'][0]
-        - not connectivity_check is success
-      with_items:
-        - kube-router
-        - flannel
-        - canal-node
-        - calico-node
-        - cilium
-
-    - name: Parse agents list
-      set_fact:
-        agents_check_result: "{{ agents.content | from_json }}"
-      delegate_to: "{{ groups['kube_control_plane'][0] }}"
+    - name: Perform netchecker tests
       run_once: true
-      when:
-        - agents is success
-        - agents.content is defined
-        - agents.content[0] == '{'
-
-    - debug:  # noqa name[missing]
-        var: agents_check_result
       delegate_to: "{{ groups['kube_control_plane'][0] }}"
-      run_once: true
-      when:
-        - agents_check_result is defined
-
-    - name: Parse connectivity check
-      set_fact:
-        connectivity_check_result: "{{ connectivity_check.content | from_json }}"
-      delegate_to: "{{ groups['kube_control_plane'][0] }}"
-      run_once: true
-      when:
-        - connectivity_check is success
-        - connectivity_check.content is defined
-        - connectivity_check.content[0] == '{'
-
-    - debug:  # noqa name[missing]
-        var: connectivity_check_result
-      delegate_to: "{{ groups['kube_control_plane'][0] }}"
-      run_once: true
-      when:
-        - connectivity_check_result is defined
+      block:
+        - name: Get netchecker agents
+          uri:
+            url: "http://{{ ansible_default_ipv4.address }}:{{ netchecker_port }}/api/v1/agents/"
+            return_content: true
+            headers:
+              Accept: application/json
+          register: agents
+          retries: 18
+          delay: "{{ agent_report_interval }}"
+          until:
+            - agents is success
+            - (agents.content | from_json | length) == (groups['k8s_cluster'] | length * 2)
+
+        - name: Check netchecker status
+          uri:
+            url: "http://{{ ansible_default_ipv4.address }}:{{ netchecker_port }}/api/v1/connectivity_check"
+            return_content: true
+            headers:
+              Accept: application/json
+          register: connectivity_check
+          retries: 3
+          delay: "{{ agent_report_interval }}"
+          until:
+            - connectivity_check is success
+            - connectivity_check.content | from_json
+
+      rescue:
+        - name: Get kube-proxy logs
+          command: "{{ bin_dir }}/kubectl -n kube-system logs -l k8s-app=kube-proxy"
+
+        - name: Get logs from other apps
+          command: "{{ bin_dir }}/kubectl -n kube-system logs -l k8s-app={{ item }} --all-containers"
+          with_items:
+            - kube-router
+            - flannel
+            - canal-node
+            - calico-node
+            - cilium
+
+        - name: Netchecker tests failed
+          fail:
+            msg: "netchecker tests failed"
 
     - name: Check connectivity with all netchecker agents
+      vars:
+        connectivity_check_result: "{{ connectivity_check.content | from_json }}"
+        agents_check_result: "{{ agents.content | from_json }}"
       assert:
         that:
           - agents_check_result is defined
@@ -193,8 +144,9 @@
             - name: samplepod
               command: ["/bin/bash", "-c", "sleep 2000000000000"]
               image: dougbtv/centos-network
+      delegate_to: groups['kube_control_plane'][0]
+      run_once: true
       when:
-        - inventory_hostname == groups['kube_control_plane'][0]
         - kube_network_plugin_multus | default(false) | bool
 
     - name: Check secondary macvlan interface
@@ -203,6 +155,7 @@
       until: output.rc == 0
       retries: 90
       changed_when: false
+      delegate_to: groups['kube_control_plane'][0]
+      run_once: true
       when:
-        - inventory_hostname == groups['kube_control_plane'][0]
         - kube_network_plugin_multus | default(false) | bool