Tonight's updates have broken startup/shutdown

[jaywalker]

It happened sometime around 22:45 GMT on my system.
I got messages like

localhost systemd[1]: Unit fedora-loadmodules.service entered failed state
localhost systemd[1]: Unit systemd-tmpfiles-clean.service entered failed state.
localhost systemd[1]: Unit plymouth-reboot.service entered failed state.

during the boot process which hangs for several minutes before eventually starting X.
On shutdown it just hangs, possibly forever - I never wait that long. :~}

I tried some half-blind poking around and discovered entries in /var/log/daemons/errors.log like this;
systemd-tmpfiles[3977]: [/usr/lib/tmpfiles.d/legacy.conf:22] Unknown group 'lock'

It seems I got the first one on 13 Jan, then every day since then, with varying frequency - possibly
because of the variable number of re-boots, possibly not. Anyway, just for the hell of it I created the
group and put root in it. The next re-boot worked fine. I haven't shut down yet 'cos I need the
machine for an overnight recording.

I confess that I have no idea what I did. I just connected one complaint about the "lock" group
with the tmp files error from systemd and thought I would give it a whirl and see what happens.

If anyone can explain what I did, or even tell me why it has nothing to do with the startup/shutdown
problem I'd be very interested.

Richard

updated PS.

The shutdown still hangs until I push the big button. Maybe it is busy doing something new,
or something that has become very difficult. Maybe.

[wobo]

Anyway, just for the hell of it I created the
group and put root in it. The next re-boot worked fine. I haven't shut down yet 'cos I need the
machine for an overnight recording.

I have exactly the same issue (plus some more). I tried your approach but it did not help anything - shutdown still hangs. I guess I will not touch this system nor the Cauldron system on my other machine (same issue) before somebody found an official solution.

[wilcal]

Ain't this fun???

[winstonteacox]

Same issue here ...

Ain't this fun???

Nope.

This is a serious issue.
I mean rock-hard serious. Dehydrated-rock-hard serious. World-class, A-1, top
of the heap, triple whopper with cheese, supersized serious. So serious that it goes way beyond the serious
we know into a whole different dimension of serious. One-of-a-kind, global, universal, intergalactic
serious. Meta-serious. Serious collapsed on itself so far that even the neutrons
have collapsed. Serious so dense that no intellect can escape. Singularly, extraordinarily, incredibly,
bewilderingly serious. Blazing hot mid-day sun on Mercury serious. Quasar serious. A monument to seriousness. I am breathless that
anything in our universe can really be this seriuos. Some pure essence of a serious so uncontaminated by anything else as to be
beyond the laws of seriousness that we know. A behemoth, a leviathan, a colossus of seriousness.


Bernd

[wilcal]

This is a serious issue.....
I mean rock-hard serious. Dehydra

I don't deny that but at this stage of the build
we're gonna encounter lots and lots of issues.
Once we get into next month or so there won't
be as many wide ranging changes so things
should settle down. This thing ain't gonna
really finalize until Aprilish.

I think the key thing to watch here are the
M2A3 Live-CD's. See how those go.

[winstonteacox]

I was kidding. Joking. Fun. etc ...

I know that Cauldron is not for the faint of heart ...


Bernd

[jaywalker]

Well I got the joke. Does that mean I've been sipping too much from the Cauldron?

As for the latest news, I got another clatter of updates tonight and I thought that many of the files were the same as the ones I got yesterday and which I am "blaming" for the breakage. So, is it fixed? Well, almost.

I took root out of the lock group - why not? When I thought a bit about it I realised that the error messages said nothing about the root user, only that the "lock" group didn't exist. The re-boot was fine and dandy and just as brisk as I have been getting to love about MGA2.

Shutdown? Not so nice. At first it looked like the same infinite hang. Then I noticed that the screen was still full of startup messages, no shutdown messages. So, I tried the three-finger salute. Two new lines popped up on vt1;

Code: Select all

Starting netprofile;   The default is the current profile.  [OK]

Checking for new hardware                                 [OK]


So I did it again and the same two lines were repeated.
So I did what any rational person would do at this point; I held down CTRL and ALT and hammered away at DELETE as fast as I could. Guess what? "Sending Sigterm to remaining processes" appeared on the screen and it started to shut down and re-boot.

RE ZULT! but .... your mileage may vary.

Richard

[winstonteacox]

Two observations:

boot:
there is a udev error message and also a error message regarding virtualbox guest additions.

shutdown:
hangs always while trying to "reload" virtualbox (absolutely unusual, so I guess there is a connection to the boot message). But Virtualbox is working flawlessly. There has been a failing shutdown already with the first 3.0... kernels, if virtualbox was installed; same symptoms as now (if I remember right) ...

Also the udev problem is something one can get used to ... just joking again.

But maybe there is a connection to virtualbox, at least regarding the shutdown ...

cheers,
Bernd

[jaywalker]

shutdown:
hangs always while trying to "reload" virtualbox

I agree - it looks most unusual to me and I don't even run virtualbox.

If you leave virtualbox out of the mix, what would it then hang at? On my system I got the distinct impression that instead of shutting down tasks for the re-boot, it was trying to repeat an initialisation it had already done. In my case it was setting the network profile.

It is almost as if runlevel 0 or 6 has not been defined or invoked correctly, but as I am still using systemd I suppose it isn't using runlevels at all! If I knew how to check how systemd tries to shutdown I would take a look myself. Ah well, that's progress.

Richard

[winstonteacox]

[....] and I don't even run virtualbox.

systemd here, too.

When I did CTRL-ALT-DEL during a failing shutdown (at "reloading" VB), the hole shutdown process started again and tried to load Virtualbox again.... very strange ... never seen something like this ...

But whatever (for the moment). Had some serious ( ) work to do, so I restored my Cauldron to the version from 13.01. .... Maybe tomorrow I will repeat the updates and will have a look again ...

cheers,
Bernd

Edit: just for clarification - I don't run Cauldron IN Virtualbox, I have a normal Cauldron with a WInXP as VB Guest ... system restore is done with a backup/restore livecd

[wobo]

I took root out of the lock group - why not? When I thought a bit about it I realised that the error messages said nothing about the root user, only that the "lock" group didn't exist. The re-boot was fine and dandy and just as brisk as I have been getting to love about MGA2.

After all updates (last was 'udisks') I still have the 3 minutes hang after "Checking for new hardware" in system start. No related message in syslog or dmesg. Then I have another break at starting net access. Takes almost 1 minute to start eth0, in Alpha2 it took not more than a couple of seconds.

Shutdown? Not so nice. So I did it again and the same two lines were repeated. So I did what any rational person would do at this point; I held down CTRL and ALT and hammered away at DELETE as fast as I could. Guess what? "Sending Sigterm to remaining processes" appeared on the screen and it started to shut down and re-boot.

Confirmed. Same messages, same hang, same crazy solution.

[jaywalker]

Confirmed. Same messages, same hang, same crazy solution.

Hey, does that mean we have ourselves a real live genuine identified and repeatable bug?

This might just be a first for me. Mostly I find I am the only one with a problem or an annoyance. But wait a minute ... we are not hearing from hordes of angry users who are now condemned to having their computers run 24/7/365.25

Perhaps when winstonteacox has put aside his foolish desire to do real work with his machine he can be persuaded to try the shutdown "fix".

In the meantime wobo, why are you still having startup problems which I have seen go away? Something about your story sounds familiar;

I still have the 3 minutes hang after "Checking for new hardware" in system start.

I used to get that, though I didn't realise it might be a limited delay rather than a hang. My "fix" for that was to CTRL-ALT-PrtScr Kill, or maybe kIllall, I don't remember now. I believe, but I don't remember clearly now, that creating the "lock" group was my permanent(?) fix for the boot issue, but I still have vt1 full of startup messages which now end with "Starting Display Manager". That is Newish. It used to clear to a login prompt as it has on Mandrake/Mandriva for hundreds of years.

Richard

[Ken-Bergen]

I have the same symptoms here.

I use ALT-(Print Screen Sysreq)-E to bypass Checking for new hardware and
ALT-(Print Screen Sysreq)-R/S/E/I/U/(O) or (B) to turn (O)ff or re(B)oot.

[wobo]

With todays 3rd update within 24 hrs of systemd (and related libraries) these problems have vanished (hopefully). The system now stops just for a few seconds at "Checking for new hardware" as usual.

[doktor5000]

I took root out of the lock group - why not? When I thought a bit about it I realised that the error messages said nothing about the root user, only that the "lock" group didn't exist. The re-boot was fine and dandy and just as brisk as I have been getting to love about MGA2.

After all updates (last was 'udisks') I still have the 3 minutes hang after "Checking for new hardware" in system start.

~3 minutes sounds like the default timeout for udev, which is 180 seconds. This is the default time it waits if some devices need to settle or if it didn't detect something properly or something along those lines. Maybe https://bugs.mageia.org/show_bug.cgi?id=3315 is related?

The lock issue is probably related to the missing update of the lockdev package, which Colin needs to look at IIUC.

[wobo]

Don't know the relation but as I wrote the problem is gone after todays updates of systemd and related libraries.

[winstonteacox]

Don't know the relation but as I wrote the problem is gone after todays updates of systemd and related libraries.

confirmed.

Now - that's fun!

cheers!

Bernd

[jaywalker]

Whoopee, just in time for me to show MGA2 to my brother tomorrow - it's a candidate to replace, or be co-resident, with the OS on his Jolibook.

Just for completeness, I tried a boot last night with no systemd and it started up and shutdown relatively easily. The biggest surprise was that it was the first time in ages that vt1 had cleared to a console login.

Richard