Can't restart/halt system with shepherd 0.9.3 after upgrading

Hey!

On a system running shepherd 0.9.3 [1], I've reconfigured, but now can't

reboot or halt.

root@hamal ~# halt

Service root is not running.

1: /gnu/store/y6w0xix15cq08qasmq75f04yzgbl98jx-shepherd-0.9.3

Hi,

Christopher Baines <mail@cbaines.net> skribis:

Hey, why halt it if it’s not running?

Seriously though, any insight from /var/log/messages? I upgraded a

bunch of machines and didn’t hit this particular problem. Bruno

reported a similar problem with 0.9.3, but this had nothing to do with

the upgrade:

https://issues.guix.gnu.org/62619

Could it be the same problem? Do you see:

Assertion (eq? (canonical-name new) (canonical-name old)) failed.

in /var/log/messages?

Ludo’.

Ludovic Courtès <ludo@gnu.org> writes:

I don't see that, but I think these are the relevant log messages:

May 24 11:17:02 localhost shepherd[1]: Evaluating user expression (and (defined? (quote transient?)) (map (# ?) ?)).

May 24 11:17:02 localhost shepherd[1]: Evaluating user expression (register-services (primitive-load "/gnu/st?") ?).

May 24 11:17:03 localhost shepherd[1]: Service host-name has been started.

May 24 11:17:03 localhost shepherd[1]: Service user-homes has been started.

May 24 11:17:03 localhost shepherd[1]: [sysctl] fs.protected_hardlinks = 1

May 24 11:17:03 localhost shepherd[1]: [sysctl] fs.protected_symlinks = 1

May 24 11:18:41 localhost shepherd[1]: Exiting shepherd...

May 24 11:18:46 localhost shepherd[1]: Grace period of 5 seconds is over; sending -337 SIGKILL.

May 24 11:23:55 localhost shepherd[1]: Service root is not running.

May 24 11:24:16 localhost last message repeated 2 times

May 24 11:30:49 localhost syslogd (GNU inetutils 2.3): restart

May 24 11:30:49 localhost vmunix: [ 0.000000] Booting Linux on physical CPU 0x0000000000 [0x410fd083]

May 24 11:30:49 localhost vmunix: [ 0.000000] Linux version 6.3.3-arm64-generic (guix@guix) (gcc (GCC) 11.3.0, GNU ld (GNU Binutils) 2.38) #1 SMP PREEMPT 1

Hi,

Christopher Baines <mail@cbaines.net> skribis:

The grace period expiration thing is probably due to the fact that

shepherd is no longer processing signals, as I described here:

https://issues.guix.gnu.org/63736

Could you share all of /var/log/messages (possibly privately, and

limiting to “shepherd” lines) starting from when the machine booted?

I’d like to see if there are hints of something that went wrong.

Ludo’.

Ludovic Courtès <ludo@gnu.org> writes:

The machine is hamal (one of the HoneyComb's) and I've added a user for

you now and added the SSH key from maintenance.git.

So you should be able to: ssh ludo@hamal.cbaines.net

Your users password is also in your home directory.

On 2023-05-24 12:27, Christopher Baines wrote:

FWIW, this has happened to me a bunch of times, I just never reported

it. Sometimes I was able to just login as root and run herd start root

to fix it.

I have an impression, from the "bunch of times" I've experienced, that

service root doesn't fail to work because of the system reconfigure, but

for some other reason.

Best regards,

David

Hi,

Christopher Baines <mail@cbaines.net> skribis:

Doesn’t work right now; anything in the logs?

Ludo’.

Ludovic Courtès <ludo@gnu.org> writes:

I believe I sorted access for Ludo, but nothing was found when looking

at the logs.

Christopher Baines <mail@cbaines.net> skribis:

I’m closing it. Let’s reopen if we stumble upon a similar issue.

Ludo’.