[systemd-devel] Too little information is shown when system enters emergency mode

[Zbigniew Jędrzejewski-Szmek]

On Sun, Oct 28, 2012 at 05:29:46PM +0000, Colin Guthrie wrote:
> 'Twas brillig, and Zbigniew Jędrzejewski-Szmek at 28/10/12 01:19 did
> gyre and gimble:
> > On Tue, Oct 23, 2012 at 04:58:46PM +0200, Lennart Poettering wrote:
> >> On Sun, 21.10.12 15:59, Andrey Borzenkov (arvidjaar at gmail.com) wrote:
> >>
> >>> Welcome to emergency mode. Use "systemctl default" or ^D to enter default mode.
> >>> Give root password for login:
> >>
> >> systemd 195 will now also mention "journalctl -b" in this
> >> message. Originally this was only in the rescue mode, because of the
> > Thanks! I tweaked the message now a bit more to take into account
> > what sulogin says by itself.
> > 
> >> assumption that if you boot directly to emergency mode then no logs
> >> would be in the journal, and hence no point in recommending this
> >> command. However, after all most of the times people will end up in
> >> emergency mode is when file systems not showing up where journald *is*
> >> actually running and includes the desired, useful information.
> >>
> >>> Started /boot/efi                                                      [  OK  ]
> >>> Dependency failed. Aborted start of /mnt                               [ ABORT]
> >>> Dependency failed. Aborted start of Login Service                      [ ABORT]
> >>> Dependency failed. Aborted start of D-Bus System Message Bus           [ ABORT]
> >>> Welcome to emergency mode. Use "systemctl default" or ^D to enter default mode.
> >>
> >> Hmm, we definitely should show the initial unit that failed in this
> >> output. 
> >>
> >> Can you restest with 195 please? If you find that there's information
> >> missing in "journalctl -b" or in the status output, then please file a
> >> bug, we really should place useful information at both.
> > It _is_ already better, the output is more complete and includes the
> > failing device, so that part is fine.
> > 
> > But there's one fundamental problem: the message suggests 'systemctl
> > default', but 'systemctl default' will fail again, unless the error
> > went away by itself, which is not going to happen in case of a missing
> > device. But it's a tough nut to crack.
> 
> Actually this might relate to a problem one of our users is encountering
> with a raid setup. The devices appear too late, but they do seem to
> appear OK (unless I'm misreading the plot output).
I think that those cases aren't that similar, since here trying to reach
default.target a second time should actually suceed.

> https://bugs.mageia.org/show_bug.cgi?id=7892
> 
> Here the user enters emergency.target and shortly after the devices seem
> to appear (I don't think they are triggered manually here).
> 
> Can anyone else read anything from the final plot attached to that bug
> and advise on how best to delay the triggering of emergency.target for a
> while? Any advice appreciated.
Hm, shouldn't the timeouts for devices that are slow to appear be increased?
Create sys-devices-...-vdaX.device with
[Unit]
JobTimeoutSec=5min
?

Zbyszek