ansible 2.9.4 - task with delegate_to: <windows_host> works, subsequent tasks fail

[Andrew Caldwell]

Hi All, hope this question makes sense. Here goes: 

Context: I have several automated patching playbooks that rely on stopping/starting services on windows before patching a database on Linux. We are using a python script to generate a dynamic inventory. We are leveraging a powershell script on a windows host to remotely start/stop services on various other windows hosts, so we only need to delegate to the one single windows host that houses the powershell scripts. We are NOT naming the windows host in the python script for dynamic inventory, and instead delegating to the FQDN of the windows host, backed by a host_vars/<windows_server>.yml for connection info for ansible to use. Oh and we are using psexec because I can't get win_shell to work to save my life.

Tasks:

- name: Pre-patching - Copy powershell script to stop service (windows)

  run_once: true

  become: false

  win_copy:

    src: ../supporting_tools/scripts/serviceshutdown{{ vmenv_result.stdout }}

    dest: C:\temp\serviceshutdown{{ vmenv_result.stdout }}.ps1

    force: no

  delegate_to: windows_server

- name: Pre-patching - Copy psexec to stop service (windows)

  run_once: true

  become: false

  win_copy:

    src: ../supporting_tools/scripts/PsExec.exe

    dest: C:\temp\PsExec.exe

    force: no

  delegate_to: windows_server

- name: Pre-patching - Stop service (windows)

  become: false

  run_once: true

  win_psexec:

    command: powershell.exe -executionpolicy bypass -noninteractive -nologo -file "C:\temp\serviceshutdown{{ vmenv_result.stdout }}.ps1"

    executable: C:\temp\PsExec.exe

    elevated: yes

    nobanner: yes

    username: "{{ansible_user}}"

    password: "{{ansible_password}}"

    interactive: no

  vars:

    ansible_become_method: runas

  delegate_to: windows_server

host_vars/windows_server.yml:

ansible_user: ad_user
ansible_password: password
ansible_connection: winrm
ansible_winrm_transport: ntlm
ansible_winrm_server_cert_validation: ignore

ansible_port: 5986

--------------------

The above role tasks run fine. Services get stopped as expected and the playbook moves on. However the next role is to gather the current repo information from the linux server that is to be patched and is a block that is a mix of delegate_to: 127.0.0.1 tasks and non delegated tasks (i.e.: they run on the target linux server). This is where the playbook falls down and goes boom. The first non-delegated task after delegating to the windows server above fails with the following error: 

fatal: [linux_server]: FAILED! => 
  msg: The powershell shell family is incompatible with the sudo become plugin

The role in question is: 

- name: Pre-patching - Update sw.repo
  block:
    - name: Pre-patching - gather minor OS target version
      shell: curl -sk -u user:pass "https://server.fqdn.com/cgi-bin/patchmm?major={{ ansible_distribution_major_version }}&os={{ os_shorthand }}"
      args:
        warn: false
      register: minor_version
      changed_when: false
      become: false
      delegate_to: 127.0.0.1
    - name: Pre-patching - gather local errata repo name
      shell: yum repolist | grep ERRATA_ | awk '{print $1}'
      register: current_errata_repo
      args:
        warn: false
      changed_when: false

  when: ansible_distribution != 'Debian' and ansible_distribution != 'Ubuntu' and not (facter_os.family == "RedHat" and ansible_distribution_major_version == "8")

The task in red is what is failing. I'm not sure why I'm getting this failure message since I'm not trying to use any powershell commands or even target a windows server. I've tried putting ansible_become_method in the host_vars file with no difference in results. Using win_shell to run a command to turn on or off a service seems to work, but I do not want to have to target 8 or 9 windows servers individually (for this one application alone. there are others with more servers behind them).

Has anyone run across this before? What is the resolution here? Is it because I am not targeting the windows host in inventory? I'd like to not do that because the entire playbook and all 50 or so included roles are geared to linux only and I do not want to have to add when clauses everywhere.

TIA,

Andrew

[Stefan Hornburg (Racke)]

On 6/11/20 7:36 PM, Andrew Caldwell wrote:
> Hi All, hope this question makes sense. Here goes: 
>
> Context: I have several automated patching playbooks that rely on stopping/starting services on windows before patching
> a database on Linux. We are using a python script to generate a dynamic inventory. We are leveraging a powershell script
> on a windows host to remotely start/stop services on various other windows hosts, so we only need to delegate to the one
> single windows host that houses the powershell scripts. We are NOT naming the windows host in the python script for
> dynamic inventory, and instead delegating to the FQDN of the windows host, backed by a host_vars/<windows_server>.yml
> for connection info for ansible to use. Oh and we are using psexec because I can't get win_shell to work to save my life.
>
> Tasks:

> *- name: Pre-patching - Copy powershell script to stop service (windows)*
> *  run_once: true*
> *  become: false*
> *  win_copy:*
> *    src: ../supporting_tools/scripts/serviceshutdown{{ vmenv_result.stdout }}*
> *    dest: C:\temp\serviceshutdown{{ vmenv_result.stdout }}.ps1*
> *    force: no*
> *  delegate_to: windows_server*
> *
> *
> *- name: Pre-patching - Copy psexec to stop service (windows)*
> *  run_once: true*
> *  become: false*
> *  win_copy:*
> *    src: ../supporting_tools/scripts/PsExec.exe*
> *    dest: C:\temp\PsExec.exe*
> *    force: no*
> *  delegate_to: windows_server*
> *
> *
> *- name: Pre-patching - Stop service (windows)*
> *  become: false*
> *  run_once: true*
> *  win_psexec:*
> *    command: powershell.exe -executionpolicy bypass -noninteractive -nologo -file "C:\temp\serviceshutdown{{
> vmenv_result.stdout }}.ps1"*
> *    executable: C:\temp\PsExec.exe*
> *    elevated: yes*
> *    nobanner: yes*
> *    username: "{{ansible_user}}"*
> *    password: "{{ansible_password}}"*
> *    interactive: no*
> *  vars:*
> *    ansible_become_method: runas*
> *  delegate_to: windows_server*
>
> host_vars/windows_server.yml:
> *ansible_user: ad_user

> ansible_password: password
> ansible_connection: winrm
> ansible_winrm_transport: ntlm
> ansible_winrm_server_cert_validation: ignore

> *
> *ansible_port: 5986*

>
>
> --------------------
> The above role tasks run fine. Services get stopped as expected and the playbook moves on. However the next role is to
> gather the current repo information from the linux server that is to be patched and is a block that is a mix of
> delegate_to: 127.0.0.1 tasks and non delegated tasks (i.e.: they run on the target linux server). This is where the
> playbook falls down and goes boom. The first non-delegated task after delegating to the windows server above fails with
> the following error: 
>
> fatal: [linux_server]: FAILED! => msg: The powershell shell family is incompatible with the sudo become plugin
>
>
> The role in question is:
>
> - name: Pre-patching - Update sw.repo block: - name: Pre-patching - gather minor OS target version shell: curl -sk -u
> user:pass "https://server.fqdn.com/cgi-bin/patchmm?major={{ ansible_distribution_major_version }}&os={{ os_shorthand }}"
> args: warn: false register: minor_version changed_when: false become: false delegate_to: 127.0.0.1 - name: Pre-patching
> - gather local errata repo name shell: yum repolist | grep ERRATA_ | awk '{print $1}' register: current_errata_repo
> args: warn: false changed_when: false
>
> when: ansible_distribution != 'Debian' and ansible_distribution != 'Ubuntu' and not (facter_os.family == "RedHat" and
> ansible_distribution_major_version == "8")
>
>
> The task in red is what is failing. I'm not sure why I'm getting this failure message since I'm not trying to use any
> powershell commands or even target a windows server. I've tried putting ansible_become_method in the host_vars file with
> no difference in results. Using win_shell to run a command to turn on or off a service seems to work, but I do not want
> to have to target 8 or 9 windows servers individually (for this one application alone. there are others with more
> servers behind them). Has anyone run across this before? What is the resolution here? Is it because I am not targeting
> the windows host in inventory? I'd like to not do that because the entire playbook and all 50 or so included roles are
> geared to linux only and I do not want to have to add when clauses everywhere.
>

Hello Andrew,

what are your connection settings for the linux_server host?

Regards
Racke

>
> TIA,
>
> Andrew
>
> --
> You received this message because you are subscribed to the Google Groups "Ansible Project" group.
> To unsubscribe from this group and stop receiving emails from it, send an email to
> ansible-proje...@googlegroups.com <mailto:ansible-proje...@googlegroups.com>.
> To view this discussion on the web visit
> https://groups.google.com/d/msgid/ansible-project/8460a481-ef9d-4811-9cf9-e6ad40255361o%40googlegroups.com
> <https://groups.google.com/d/msgid/ansible-project/8460a481-ef9d-4811-9cf9-e6ad40255361o%40googlegroups.com?utm_medium=email&utm_source=footer>.


--
Ecommerce and Linux consulting + Perl and web application programming.
Debian and Sympa administration. Provisioning with Ansible.

[Andrew Caldwell]

@Racke,

These are my ansible.cfg settings. The playbook is called with 'become: true' since most tasks require elevated rights:

[defaults]

forks          = 60

poll_interval  = 15

become_user    = root

remote_port    = 22

host_key_checking = False

become_exe     = sudo

timeout        = 60

callback_whitelist = timer, profile_tasks

display_skipped_hosts = no

force_color = True

stdout_callback = yaml

bin_ansible_callbacks = True

gather_subset=!all,min,facter,!hardware,!network,!virtual,!ohai

[paramiko_connection]

record_host_keys = False

host_key_checking = False

[ssh_connection]

ssh_args = -o ControlMaster=no -o ControlPersist=60s -o ControlPath=/tmp/ssh-%h -o StrictHostKeyChecking=no

scp_if_ssh = True

retries = 2

pipelining = True

[Andrew Caldwell]

In hindsight this is pretty obvious -- I did not realize I was stepping on some ansible built-ins like become_method and whatnot. adding a set_fact task after the windows execution solved this issue...