That makes some sort of sense.
The action for reload is "Just" a signal to MAINPID
ExecReload=/bin/kill -HUP $MAINPID
I think the failure is a second -HUP signal before the former reload is fully complete. And on a more complex keepalived setup this might also be a longer time-window to trigger.
Yet the service files in Xenial and Zesty are identical, and I'd not know one can wait for a signal to be handled.
Might even be a systemd fix in Zesty that now "waits" for such to complete.
So lessons learned, I seem to be able to reliably trigger it with a head to head restart:
"sudo systemctl restart keepalived; sudo systemctl restart keepalived;"
That makes some sort of sense. /bin/kill -HUP $MAINPID
The action for reload is "Just" a signal to MAINPID
ExecReload=
I think the failure is a second -HUP signal before the former reload is fully complete. And on a more complex keepalived setup this might also be a longer time-window to trigger.
Yet the service files in Xenial and Zesty are identical, and I'd not know one can wait for a signal to be handled.
Might even be a systemd fix in Zesty that now "waits" for such to complete.
So lessons learned, I seem to be able to reliably trigger it with a head to head restart:
"sudo systemctl restart keepalived; sudo systemctl restart keepalived;"
Going to my Zesty env to confirm that there.