Fix Podman "cgroupv2 manager is set to systemd but there is no systemd user session"
Checked PAM locally: /etc/pam.d/su and su-l (util-linux) have no pam_systemd line, and /etc/pam.d/sudo uses pam_systemd.so class=none, which pam_systemd(8) documents as skipping logind registration and not starting the user manager (v258+). So neither su nor sudo -u gives XDG_RUNTIME_DIR or a user manager, as the cause says. machinectl is in the systemd package (261.2-1). rootlesscontaine.rs says exactly this and recommends machinectl shell user@ or ssh user@localhost, plus enable-linger. The Arch wiki Podman page supports lingering for containers that stop at logout. podman issue #17202 has the exact warning title. Note that sshd is off by default on Omarchy, so the ssh alternative needs it enabled first, but machinectl is listed first and works. Not exercised: podman is not installed here.
Symptom
Every rootless podman command prints:
The cgroupv2 manager is set to systemd but there is no systemd user session available. For using systemd, you may need to login using an user session. Alternatively, you can enable lingering with: `loginctl enable-linger 10003` (possibly as root). Falling back to --cgroup-manager=cgroupfs
I got into the account with su - user or sudo -u user. --cpus and --memory limits do nothing, and containers stop when I log out.
Cause
Rootless podman talks to the user's systemd instance over $XDG_RUNTIME_DIR and the user D-Bus to place containers in cgroups. Switching from root with su -l or sudo -u does not create a logind session, so XDG_RUNTIME_DIR is unset and no user manager is reachable, and podman falls back to cgroupfs, which has no delegated controllers for resource limits.
Fix
Get a real login session for that user instead of su/sudo:
sudo machinectl shell user@ # machinectl ships in Arch's systemd package
# or
ssh user@localhost
For a service account whose containers must keep running with nobody logged in, enable lingering so its user manager starts at boot:
sudo loginctl enable-linger user
If you must keep using su, lingering first, then point the shell at the runtime directory:
export XDG_RUNTIME_DIR=/run/user/$(id -u)
Silencing it with cgroup_manager = "cgroupfs" under [engine] in ~/.config/containers/containers.conf removes the warning but keeps the limitation.
Verify
echo $XDG_RUNTIME_DIR prints /run/user/<uid>, systemctl --user status answers, podman info --format '{{.Host.CgroupManager}}' prints systemd, and the warning is gone.