[systemd-devel] Reboot hangs at the very last step

[Andrei Borzenkov]

18.07.2016 17:35, Frank Steiner пишет:
> Hi,
> 
> on some of our servers (all of those are NFS-servers with iscsi devices...)
> rebooting sometimes hangs after
> 
> [  OK  ] Reached target Shutdown.
> 
> The next line, the kernel mesage about resetting the system, never
> shows up. Is there a way to figure out what systemd is waiting for 
> at this point?
> 

In the past this often was associated with network being shut down too
early.

> I tried the debug shell with a while-true loop doing "systemctl list-jobs",
> but the last output I get before the debug shell itself hangs looks quite 
> ok (see screenshot).
> 
> On one server I can reliably trigger the hang: I removed iscsi.service
> so the iscsi-mounts in /etc/fstab timeo ut during the boot. After the
> server has come up, I manually activate the iscsi-devices with iscsiadm,
> then run "vgchange -ay", then reboot the server, and it hangs.
> Without manually activating the devices or if iscsi.service is there 
> during boot, the server will reboot fine.
> 

Well, that rather enforces this; when network filesystem is part of
system startup it gets extra dependency on network and is unmounted
before network is stopped; but manual mount does not (or at least did
not - I believe there were some patches) have this dependency so may be
delayed.

Try to closely look at all units that need network - do all of them have
proper Before and After to ensure they are stopped before network?

> Unfortunately, when I add  "systemd.log_level=debug" the server will
> even reboot in the scenario that otherwise triggers the hang :-(
> 
> cu,
> Frank
> 
> 
> 
> 
> 
> _______________________________________________
> systemd-devel mailing list
> systemd-devel at lists.freedesktop.org
> https://lists.freedesktop.org/mailman/listinfo/systemd-devel
>