[BUG] Windows agent service doesn't bind to port 6556 until restarted after installation

[zenwalk2013]

Describe the bug
The CheckMK Windows agent service starts successfully after installation/registration but does not bind to port 6556 until the service is restarted. This causes the win_wait_for port verification task in the agent role to timeout, even though the service is running.
After installation completes:
CheckmkService status shows as "Running"
Port 6556 is NOT listening (verified with netstat -an | findstr 6556)
After manually restarting the service with restart-service CheckmkService, port 6556 becomes active and listening

This appears to be a Windows-specific issue where the agent controller (cmk-agent-ctl.exe) doesn't immediately initialize the network listener after installation/registration, requiring a service restart to fully activate.

Component Name
Component Name: ansible_collections/checkmk/general/roles

Ansible Version
ansible [core 2.20.0]
jinja version = 3.1.6
pyyaml version = 6.0.3 (with libyaml v0.2.5)

**Checkmk Version and Edition**
CheckMK version: 2.4.0p3

**Collection Version**
checkmk.general                          6.5.0

To Reproduce
Steps to reproduce the behavior:

Install CheckMK agent on Windows 11 host using the checkmk.general.agent role
Use configuration with checkmk_agent_mode: 'pull' and checkmk_agent_tls: false
Wait for the role to complete agent installation and registration
Observe the win_wait_for task timeout with error:

TASK [checkmk.general.agent : Win32NT: Verify Checkmk Agent Port is open.] ****
fatal: [monitoring-host: FAILED! => {
"changed": false,
"elapsed": 60.900359599999994,
"msg": "timeout while waiting for 127.0.0.1:6556 to start listening",
"wait_attempts": 20
}

Expected behavior
After the CheckMK agent installation and registration completes:

The CheckmkService should be running AND listening on port 6556
The win_wait_for task should successfully verify the port is open
No manual service restart should be required

Actual behavior
After installation completes:The CheckmkService shows as "Running" but is NOT listening on port 6556The win_wait_for task times out after 60 secondsManual service restart is required to bind port 6556PS C:\Users\user> get-service CheckmkServiceStatus Name DisplayName------ ---- -----------Running CheckmkService Checkmk ServicePS C:\Users\user> netstat -an | findstr 6556# No output - port not listening despite service runningPS C:\Users\user> restart-service CheckmkServicePS C:\Users\user> netstat -an | findstr 6556TCP 0.0.0.0:6556 0.0.0.0:0 LISTENINGTCP [::]:6556 [::]:0 LISTENING

Minimum reproduction example

• name: Install CheckMk agent from monitoring server
  ansible.builtin.include_role:
  name: checkmk.general.agent
  vars:
  checkmk_agent_version: "2.4.0p3"
  checkmk_agent_server_protocol: https
  checkmk_agent_server_validate_certs: false
  checkmk_agent_server_port: 443
  checkmk_agent_configure_firewall: false
  checkmk_agent_site: 'sci_monitoring'
  checkmk_agent_user: "{{ checkmk_automation_user }}"
  checkmk_agent_secret: "{{ checkmk_automation_secret }}"
  checkmk_agent_registration_server_protocol: "https"
  checkmk_agent_add_host: false
  checkmk_agent_host_name: "{{ inventory_hostname }}"
  checkmk_agent_tls: false
  checkmk_agent_mode: 'pull'

Additional context
workaround by adding restart plays after installation play in the playbook

• name: Install CheckMk agent from monitoring server
  ansible.builtin.include_role:
  name: checkmk.general.agent
  vars:
  checkmk_agent_mode: 'ssh' # Skip port check to avoid timeout

• name: Restart CheckMK service to ensure port binding (Windows)
  ansible.windows.win_service:
  name: CheckMkService
  state: restarted
  when: ansible_os_family == "Windows"

• name: Wait for CheckMK agent port to be listening (Windows)
  ansible.windows.win_wait_for:
  port: 6556
  timeout: 30
  when: ansible_os_family == "Windows"

Suggested fix
The agent role's Win32NT.yml tasks should include an automatic service restart after agent installation/registration and before the port verification check. This would ensure the agent is fully initialized and listening on the correct port.
Proposed change location: roles/agent/tasks/Win32NT.yml - add a service restart task before line 123 (the win_wait_for port verification task).

OS: Windows 11 (also reported on Windows Server 2019/2022)
Issue occurs: On both fresh installations and updates
Affected versions:

Previously observed with collection version 5.10.1
Still present in version 6.5.0

Agent mode: pull mode without TLS
This is a Windows-specific issue - Linux agent installations work correctly

[robin-checkmk]

@zenwalk2013 thanks for reporting this! Initial thoughts: This sounds more like a Checkmk bug, than something wrong with the collection. I would like to avoid blanket-restart-tasks if possible. So let me ask you this: Did you also test this on different Checkmk versions? E.g., 2.3.0 or with an up-to-date 2.4.0? Is there any snake oil (also known as antivirus or endpoint protection these days) on the affected host(s)?

I only have very limited means to test and verify this on my end and I would assume, that this is not a general issue. So please bear with me, while we try to understand this.

[zenwalk2013]

Hi, robin-checkmk. Thanks for your response.

Only checkmk version I had this issue is 2.4.0p3 on server side. I havent tested with other versions yet. I see few adjacent issue on checkmk forum confirming repetitive windows checkmk agent bug. There are plans to move to next stable release soon. I will update here with results

https://forum.checkmk.com/t/windows-cmk-agent-ctl-not-binding-to-port-6556-sometimes/32659
https://forum.checkmk.com/t/checkmk-agent-on-windows-systems-stop-listening-on-port-6556/38949
https://forum.checkmk.com/t/do-you-have-checkmk-agent-2-4-sucessful-running-on-win11-24h2-all-updates/54477

Multiple users confirmed downgrading to 2.3.x resolves the issue.

In relation to the hosts, all windows i.e. 46 stations that were added to monitoring site (mostly hardware but some are vms) had this issue. I had to restart checkmk service for all to have 6556 tcp port in listening state.

[robin-checkmk]

@zenwalk2013 thanks for the feedback. Then please verify, that this issue persists on a current Checkmk 2.4.0, as the forum post suggests, that it is a Checkmk issue and not related to the collection.