kubespray/scripts/collect-info.yaml

---
- hosts: all
  become: true
  gather_facts: no

  vars:
    docker_bin_dir: /usr/bin
    bin_dir: /usr/local/bin
    ansible_ssh_pipelining: true
    etcd_cert_dir: /etc/ssl/etcd/ssl
    kube_network_plugin: calico
    archive_dirname: collect-info
    commands:
      - name: timedate_info
        cmd: timedatectl status
      - name: kernel_info
        cmd: uname -r
      - name: docker_info
        cmd: "{{ docker_bin_dir }}/docker info"
      - name: ip_info
        cmd: ip -4 -o a
      - name: route_info
        cmd: ip ro
      - name: proc_info
        cmd: ps auxf | grep -v ]$
      - name: systemctl_failed_info
        cmd: systemctl --state=failed --no-pager
      - name: k8s_info
        cmd: "{{ bin_dir }}/kubectl get all --all-namespaces -o wide"
      - name: errors_info
        cmd: journalctl -p err --no-pager
      - name: etcd_info
        cmd: "{{ bin_dir }}/etcdctl --peers={{ etcd_access_addresses | default('http://127.0.0.1:2379') }} cluster-health"
      - name: calico_info
        cmd: "{{ bin_dir }}/calicoctl node status"
        when: '{{ kube_network_plugin == "calico" }}'
      - name: calico_workload_info
        cmd: "{{ bin_dir }}/calicoctl get workloadEndpoint -o wide"
        when: '{{ kube_network_plugin == "calico" }}'
      - name: calico_pool_info
        cmd: "{{ bin_dir }}/calicoctl get ippool -o wide"
        when: '{{ kube_network_plugin == "calico" }}'
      - name: weave_info
        cmd: weave report
        when: '{{ kube_network_plugin == "weave" }}'
      - name: weave_logs
        cmd: "{{ docker_bin_dir }}/docker logs weave"
        when: '{{ kube_network_plugin == "weave" }}'
      - name: kube_describe_all
        cmd: "{{ bin_dir }}/kubectl describe all --all-namespaces"
      - name: kube_describe_nodes
        cmd: "{{ bin_dir }}/kubectl describe nodes"
      - name: kubelet_logs
        cmd: journalctl -u kubelet --no-pager
      - name: coredns_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l k8s-app=coredns -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system; done"
      - name: apiserver_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l component=kube-apiserver -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system; done"
      - name: controller_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l component=kube-controller-manager -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system; done"
      - name: scheduler_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l component=kube-scheduler -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system; done"
      - name: proxy_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l k8s-app=kube-proxy -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system; done"
      - name: nginx_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l k8s-app=kube-nginx -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system; done"
      - name: flannel_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l app=flannel -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system flannel-container; done"
        when: '{{ kube_network_plugin == "flannel" }}'
      - name: canal_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l k8s-app=canal-node -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system flannel; done"
        when: '{{ kube_network_plugin == "canal" }}'
      - name: calico_policy_logs
        cmd: "for i in `{{ bin_dir }}/kubectl get pods -n kube-system -l k8s-app=calico-kube-controllers -o jsonpath={.items..metadata.name}`;
          do {{ bin_dir }}/kubectl logs ${i} -n kube-system ; done"
        when: '{{ kube_network_plugin in ["canal", "calico"] }}'
      - name: helm_show_releases_history
        cmd: "for i in `{{ bin_dir }}/helm list -q`; do {{ bin_dir }}/helm history ${i} --col-width=0; done"
        when: "{{ helm_enabled|default(true) }}"

    logs:
      - /var/log/syslog
      - /var/log/daemon.log
      - /var/log/kern.log
      - /var/log/dpkg.log
      - /var/log/apt/history.log
      - /var/log/yum.log
      - /var/log/messages
      - /var/log/dmesg

  environment:
    ETCDCTL_CERT_FILE: "{{ etcd_cert_dir }}/node-{{ inventory_hostname }}.pem"
    ETCDCTL_KEY_FILE: "{{ etcd_cert_dir }}/node-{{ inventory_hostname }}-key.pem"

  tasks:
    - name: set etcd_access_addresses
      set_fact:
        etcd_access_addresses: |-
          {% for item in groups['etcd'] -%}
            https://{{ item }}:2379{% if not loop.last %},{% endif %}
          {%- endfor %}          
      when: "'etcd' in groups"

    - name: Storing commands output
      shell: "{{ item.cmd }} 2>&1 | tee {{ item.name }}"
      failed_when: false
      with_items: "{{ commands }}"
      when: item.when | default(True)
      no_log: True

    - name: Fetch results
      fetch: src={{ item.name }} dest=/tmp/{{ archive_dirname }}/commands
      with_items: "{{ commands }}"
      when: item.when | default(True)
      failed_when: false

    - name: Fetch logs
      fetch: src={{ item }} dest=/tmp/{{ archive_dirname }}/logs
      with_items: "{{ logs }}"
      failed_when: false

    - name: Pack results and logs
      archive:
        path: "/tmp/{{ archive_dirname }}"
        dest: "{{ dir|default('.') }}/logs.tar.gz"
        remove: true
      delegate_to: localhost
      connection: local
      become: false
      run_once: true

    - name: Clean up collected command outputs
      file: path={{ item.name }} state=absent
      with_items: "{{ commands }}"