Fails to register machines on machined.
kalium99 opened this issue · 2 comments
Running kube-spawn start fails saying it timed out waiting for machines to start.
Poking around a bit, I can see that systemd-nspawn is in fact running, and I can see the scope is running in systemd. However, machinectl does not list any machines.
If I copy and paste the systemd-nspawn command output from ps, then the container starts fine and is registered with systemd-machined
However it seems that when you run /usr/bin/systemd-run --scope --property=DevicePolicy=auto /usr/bin/kube-spawn cni-spawn ... command that then forks the systemd-nspawn, it just fails to register with systemd-machined. As far as I can tell the container is running...
I can't see anything useful in the logs either.
Has anyone seen these symptoms before?
This seems like a different problem to #339
I'm running systemd v243 on Fedora 31.
So looking further, it seems to be a problem with the way that systemd-nspawn is being launched.
For example when running systemd-nspawn directly under systemd-run it works fine, but when it's being ForkExec'ed from kube-spawn it doesn't.
Looking at the dbus-monitor --system and it's clear that there is nothing being sent to the org.freedesktop.machine1.Manager which explains why nothing is being registered with systemd-machined.
Still making my way through the strace to see what I can see, but this looks like it's likely a bug (or feature?) in systemd and nothing to do with kube-spawn.
Running into this issue as well, it does not work on ArchLinux.
Is this project still maintained ?