systemctl start fails causing fail on ansible run about rke2-ansible HOT 3 CLOSED

@jberger Could you elaborate on the conditions required to observe the failure and which type of nodes you're seeing this for? We have tried to work in enough staggering logic where each "start" happens in sequence and the systemd unit then waits until ETCD has joined to return exit 0.

from rke2-ansible.

I don't think this is related to staggering. Even when doing the deploy manually (without the ansible stuff) sometimes the initial systemctl start rke2-server.service exits 1. It then continues to retry until it works. Mostly I think this is due to round-robin DNS for the registration url. Anyway, the logic later on in ansible seems to be fine for detecting that the kubelet and etcd come up it just is noise that the start shows as an error.

Anyway, if you need any other details I'm happy to give them. Ubuntu 20.04, rke2 latest. Baremetal servers.

from rke2-ansible.

It would be good not to expect systemd here at all. systemd is 1m+ lines of code, not really fit for purpose on a small lean server that just wants to run rke. For these and other reasons, Devuan (debian/ubuntu compatible in all other ways) doesn't include it, and instead provides options of sysvinit, runit, etc.

from rke2-ansible.