[systemd-devel] Very long delay (~60 seconds) when stopping or restarting systemd-nspawn containers

[Gena Makhomed]

Hello, All!

CentOS 8.2, systemd 239, systemd-networkd 245 from epel repo.

containers created with systemd-nspawn don't stop as expected.

command "machinectl stop $CONTAINER" outside container,
or command "poweroff" inside container wait ~60 seconds
in "zombie" state before finally stopping container.

"zombie" state means what inside container still running
only these 3 processes:

/usr/lib/systemd/systemd
/usr/lib/systemd/systemd-journald
/usr/lib/systemd/systemd-networkd

in log file of container I can see:

Jun 25 13:21:48 core1-router systemd[1]: Received SIGRTMIN+4.
Jun 25 13:22:50 core1-router systemd[1]: user at 0.service: Processes still 
around after final SIGKILL. Entering failed mode.
Jun 25 13:22:50 core1-router systemd[1]: user at 0.service: Failed with 
result 'timeout'.
Jun 25 13:22:50 core1-router systemd[1]: Stopped User Manager for UID 0.

How I can debug/workaround or solve this problem with very long delay
when containers stop/restart ?

On hardware node:

/etc/systemd/nspawn/1.nspawn

[Exec]
ResolvConf=copy-host
LimitNOFILE=infinity
LimitNICE=40

[Network]
Bridge=venet0


Inside container:

/etc/systemd/network/00-host0.network

[Match]
Name=host0

[Network]
Address=172.16.1.1/24
Gateway=172.16.1.254

-- 
Best regards,
  Gena