`skupper.v2.system` with action `reload` on systemd sites will not restart the `skrouterd` process

[hash-d]

Reproducer TBD:

• Create a systemd site
• Run a skupper.v2.system task with action reload
• ps -fC skrouterd ⟶ the process has not been restarted.

Skupper 2.1.4
Ansible collection skupper.v2 2.1.3

[hash-d]

Reproducer created on https://gist.github.com/hash-d/3dd66e5d10dd262ad91336df97575445

[hash-d]

If you run that playbook completely, and then manually kill the skrouterd process and re-run the playbook, it fails:

$ ansible-playbook -i /tmp/inv.yaml 95.yaml 

PLAY [Reproducer for https://github.com/skupperproject/skupper-ansible/issues/95] ***********************

TASK [Create site] **************************************************************************************
ok: [some_host]

TASK [Start it] *****************************************************************************************
[WARNING]: namespace 'issue-95' already exists
ok: [some_host]

TASK [Run ps before reload] *****************************************************************************
fatal: [some_host]: FAILED! => {"changed": true, "cmd": ["ps", "-fC", "skrouterd"], "delta": "0:00:00.015140", "end": "2026-05-06 14:34:58.856761", "msg": "non-zero return code", "rc": 1, "start": "2026-05-06 14:34:58.841621", "stderr": "", "stderr_lines": [], "stdout": "UID          PID    PPID  C STIME TTY          TIME CMD", "stdout_lines": ["UID          PID    PPID  C STIME TTY          TIME CMD"]}

PLAY RECAP **********************************************************************************************
some_host                : ok=2    changed=0    unreachable=0    failed=1    skipped=0    rescued=0    ignored=0   

Note that on the start action, it just says that the namespace already exists, and does not try to actually start it.

This is somewhat similar to #96, but here it is on skrouterd whereas there it is on the controller.

Let me know if you want a new ticket for this one.

[hash-d]

Note this second issue above (manual kill) happens for podman sites, too: after a podman rm -f on the container, action: start does not restart it.