update_pgcluster.yml: Reboot node if it's required, e.g. kernel or security updates

[chuegel]

Environments like PCIDSS require to apply kernel and security updates on a regular basis. Once those updates have been applied, a reboot of the host is needed. This PR will reboot a host only if it's required (kernel updates, security updates etc.) when updating the PostgreSQL cluster with ansible-playbook update_pgcluster.yml -e target=system

[chuegel]

The failing tests are expected as the docker containers die on reboot. Not sure how to catch this test. Vagrant?

What worries me is, as I mentioned in the discussion the handler get called multiple times when notified. This would need a thorough testing.

[vitabaks]

Thank you @chuegel for your contribution!

The failing tests are expected as the docker containers die on reboot.

You can exclude this task from containers

ansible_virtualization_type not in ['container', 'docker', 'lxc', 'podman']  # exclude for containers to prevent test failures in CI.

[vitabaks]

What worries me is, as I mentioned in the discussion the handler get called multiple times when notified.

perhaps it is necessary to move from "handler" to a simple task.

[vitabaks]

perhaps it is necessary to move from "handler" to a simple task.

Example:

- name: Check if a reboot is required
  ansible.builtin.stat:
    path: /var/run/reboot-required
  register: reboot_required_debian
  changed_when: false
  when:
    - ansible_os_family == "Debian"
    - ansible_virtualization_type not in ['container', 'docker', 'lxc', 'podman']  # exclude for containers to prevent test failures in CI.

- name: Check if a reboot is required
  ansible.builtin.command: needs-restarting -r
  register: reboot_required_rhel
  failed_when: false
  changed_when: false
  when:
    - ansible_os_family == "RedHat"
    - ansible_virtualization_type not in ['container', 'docker', 'lxc', 'podman']  # exclude for containers to prevent test failures in CI.

- name: Rebooting host
  ansible.builtin.reboot:
    msg: "Reboot initiated by Ansible due to required system updates"
    reboot_timeout: 1800 # 30 minutes
    test_command: uptime
  when: (reboot_required_debian.stat.exists is defined and reboot_required_debian.stat.exists) or
        (reboot_required_rhel.rc is defined and reboot_required_rhel.rc != 0)

[chuegel]

perhaps it is necessary to move from "handler" to a simple task.

Example:

- name: Check if a reboot is required
  ansible.builtin.stat:
    path: /var/run/reboot-required
  register: reboot_required_debian
  changed_when: false
  when:
    - ansible_os_family == "Debian"
    - ansible_virtualization_type not in ['container', 'docker', 'lxc', 'podman']  # exclude for containers to prevent test failures in CI.

- name: Check if a reboot is required
  ansible.builtin.command: needs-restarting -r
  register: reboot_required_rhel
  failed_when: false
  changed_when: false
  when:
    - ansible_os_family == "RedHat"
    - ansible_virtualization_type not in ['container', 'docker', 'lxc', 'podman']  # exclude for containers to prevent test failures in CI.

- name: Rebooting host
  ansible.builtin.reboot:
    msg: "Reboot initiated by Ansible due to required system updates"
    reboot_timeout: 1800 # 30 minutes
    test_command: uptime
  when: (reboot_required_debian.stat.exists is defined and reboot_required_debian.stat.exists) or
        (reboot_required_rhel.rc is defined and reboot_required_rhel.rc != 0)

Yeah, that looks good, thank you. I tested it with an 3 node Ubuntu 22.04 cluster and it works as expected.