If you run containerd as a systemd service, and you try to restart the service while containers are running the systemctl restart containerd will block because while containerd (the daemon) exits when it gets a SIGTERM it doesn't attempt to kill any of the containerd-shims -- which then causes systemd to SIGKILL them after 10 seconds (or whatever the timeout is).
If you run
containerdas a systemd service, and you try to restart the service while containers are running thesystemctl restart containerdwill block because whilecontainerd(the daemon) exits when it gets aSIGTERMit doesn't attempt to kill any of thecontainerd-shims -- which then causes systemd toSIGKILLthem after 10 seconds (or whatever the timeout is).