How to shutdown when systemd crashes?

clivem · #1 3rd May 2011, 05:12 PM

How do you cleanly shutdown following a systemd crash?

Several times over the last few days systemd has failed. Didn't even know it had crashed until it was time to shutdown or reboot and the usual "shutdown -r/-h now" wouldn't work.

Code:

[clivem@c7p6t7ws ~]$ sudo shutdown -r now
Failed to open /dev/initctl: No such device or address
Failed to talk to init daemon.

"systemctl reboot" doesn't work either

Code:

[clivem@c7p6t7ws ~]$ sudo systemctl reboot
Failed to get D-Bus connection: Failed to connect to socket /org/freedesktop/systemd1/private: Connection refused

ryptyde · #2 4th May 2011, 01:35 AM

Quote:

Originally Posted by clivem

How do you cleanly shutdown following a systemd crash?

Several times over the last few days systemd has failed. Didn't even know it had crashed until it was time to shutdown or reboot and the usual "shutdown -r/-h now" wouldn't work.

Code:

[clivem@c7p6t7ws ~]$ sudo shutdown -r now
Failed to open /dev/initctl: No such device or address
Failed to talk to init daemon.

"systemctl reboot" doesn't work either

Code:

[clivem@c7p6t7ws ~]$ sudo systemctl reboot
Failed to get D-Bus connection: Failed to connect to socket /org/freedesktop/systemd1/private: Connection refused

I use su instead of sudo then:

Code:

shutdown -h now

phil

jpollard · #3 4th May 2011, 03:20 AM

even that depends on init (now systemd) for proper closeout, terminating services, and dismounting filesystems to prevent corruption (though that happens less now).

shutdown should just hang as well.

pankajp · #4 4th May 2011, 10:09 AM

Though i'm not sure how to shutdown in case of systemd crash, you should update to version 26 of systemd from koji/updates-testing. That fixes the systemd crash due to use of systemadm. (I was suffereing from same problem, and had to hard-shutdown using power button)

jpollard · #5 4th May 2011, 10:36 AM

Actually, I'm surprised the system didn't panic.

A panic has been the normal thing for the kernel to do when init fails.

I suspect it didn't fail, but hung, or stopped responding to requests. This used to be a simple thing - a signal. It is now a network level request that requires the service to spawn a daemon, and that introduces more complex demands for reliability.