Boot failed to complete after fsck

At the time when it appeared stuck, there was still heavy disk access. Looking at the sysrq trace and at the superblock on my third local filesystem, it seems like another fsck was running without displaying any output on the console.

This is the state the system was left in after the control+alt+del

This problem is actively reporter by manu users here (http://ubuntuforums.org/showthread.php?t=36504) and here (http://ubuntuforums.org/showthread.php?t=245216)

My gdm restarts too.
Possible reasons:
* random screensaver
Now it's blank only, was "random" every time.
But gdm restarted while active keyboard work :), it seems not to be SS-related
* anacrod/crond/atd
* ipv6?

Distro is Ubuntu Edgy 6.10

There are no random restarts mentioned by the original poster, so I don't think the threads that were linked are relevant.

Matt: can you provide your fstab as well? The ordering and listings of fsck might be useful.

On Fri, Sep 07, 2007 at 12:38:44PM -0000, David Tomaschik wrote:
> There are no random restarts mentioned by the original poster, so I
> don't think the threads that were linked are relevant.
>
> Matt: can you provide your fstab as well? The ordering and listings of
> fsck might be useful.

#
# <file system> <mount point> <type> <options> <dump> <pass>
# /dev/sda1 -- converted during upgrade to edgy
UUID=daf1bf61-d974-4e90-b2c7-5f4dfa4e56cf / ext3 errors=remount-ro 0 1
proc /proc proc defaults 0 0
tmpfs /tmp tmpfs nosuid,nodev 0 0
# /dev/sda5 -- converted during upgrade to edgy
UUID=1893b08b-4da1-47b7-b4d8-ba31f5cbe9c4 /home ext3 errors=remount-ro,nosuid,nodev 0 2
UUID=62373d5a-e830-4d96-8c8b-33a26d83a45c swap swap defaults 0 0
UUID=3cab53b1-68e8-431e-938e-8fc2ab76b334 /space ext3 rw,nosuid,nodev 0 2

--
 - mdz

Just to add another data point, I've just seen the same, or a very similar problem with Hardy alpha (up to date as of March 6, 2008). And I saw it another time somewhere in the last month.

What I just saw: I get a fsck at boot on an ext3 fs that hasn't been checked in a while (so it's not an fs with errors), it displays the progress on the boot splash, but hangs at 83%. (The display of progress on the boot splash seems to be a recently added feature.) When I switch to the console, I see output that seems to indicate that the fsck has actually finished, however, there's no output after the fsck, and the boot does not continue. There's no more disk activity either. Pressing Ctrl+Alt+Del at that point actually seems to make the boot continue. I get various errors which seem to indicate that the root fs is still mounted read-only though. X won't start either. It requires a second Ctrl+Alt+Del to start a reboot, which then works just fine.

My fstab records, I think the fsck was on sda5 (mounted at /otheros):

proc /proc proc defaults 0 0
# /dev/sda2
UUID=b2905f36-9a1a-4199-b2f8-8f527cef7eef / ext3 errors=remount-ro,relatime 0 1
# /dev/sda6
UUID=c9d56401-32c0-42c8-8eec-28bb0d0c240d /data ext3 defaults,relatime 0 2
# /dev/sda5
UUID=ec1b5cb8-c8b7-4c69-bf6e-734b75f2f605 /otheros ext3 defaults,relatime 0 2
# /dev/sda1
UUID=8048A4E648A4DBE2 /windows ntfs defaults,umask=007,gid=46 0 1
# /dev/sda3
UUID=588534c7-7b82-4129-abd1-cf16bd1963b7 none swap sw 0 0
/dev/scd0 /media/cdrom0 udf,iso9660 user,noauto,exec 0 0

# NAS filesystems
//nas/muziek /nas/muziek smbfs credentials=...,uid=1000,gid=1000 0 0
//nas/backups /nas/backups smbfs credentials=...,uid=1000,gid=1000 0 0
//nas/net /nas/net smbfs credentials=...,uid=1000,gid=1000,noauto 0 0

Bart, recently a few bugs in usplash have been fixed which caused hiccups like those. Do you still get the problem on current hardy? (With usplash version 0.5.16)

Martin Pitt wrote:
> Bart, recently a few bugs in usplash have been fixed which caused
> hiccups like those. Do you still get the problem on current hardy? (With
> usplash version 0.5.16)

I'll keep an eye out for it. I have my file systems set to check every
30 days, so I'll expect one to trigger soon.

Cheers,
Bart

Just for the record, you can trigger that routine check with
"sudo tune2fs -C 30 /dev/sda1" (or whichever partition you mount by
default, see output of "mount").

Just had a spontaneous one. Worked like a charm, problem is gone! Did have a minor nit, but that's a separate issue which I reported as #203711.

I use Hardy beta and still have the symptoms described earlier:

* fsck check after certain amount of boots
* see only text (only console) with no status bar, but hear hard disk activity
* after a while hard disk activity is gone, but no progress
* try to ctrl-c - nothing
* try to ctrl-alt-del - surprisingly X starts, but the checked partition is not mounted

rubinstein, can you please check this again on latest hardy? I think all cases of broken fsck usplash integration are fixed now.

OK, Hardy with latest updates.

* usplash 640x480
* when checking filesystem: usplash gone, console shows fsck working as it counts to 100%
* GDM starts

So now fsck works as intended and completes a check, but I still see no graphical progress bar as it changes to console mode.

rubinstein [2008-03-31 18:07 -0000]:
> * when checking filesystem: usplash gone, console shows fsck working as it counts to 100%
>
> So now fsck works as intended and completes a check, but I still see no
> graphical progress bar as it changes to console mode.

Graphical progress bar in text mode currently does not work with
current fsck, since we cannot switch on the fly. The remaining bug is
that usplash terminates early in the first place.

I've got a similar issue with my mythbuntu (jaunty) box. At startup the usplash "freezes" at about 10 % without displaying that it is doing a fsck. After 20 minutes I switched to console 8 to see what's up. fsck stood at 63 %. After a while it finished without showing where it went from 63 % to 100 %. I did a reboot and it came up as there was no matter before. I deactivated the routine check, because my wife or my daughter don't know what to do next time when I'm not around.

Description: Ubuntu 9.04
Release: 9.04
upstart:
  Installiert: 0.3.9-8
  Kandidat: 0.3.9-8
  Versions-Tabelle:
 *** 0.3.9-8 0
        500 http://de.archive.ubuntu.com jaunty/main Packages
        100 /var/lib/dpkg/status
usplash:
  Installiert: 0.5.31
  Kandidat: 0.5.31
  Versions-Tabelle:
 *** 0.5.31 0
        500 http://de.archive.ubuntu.com jaunty/main Packages
        100 /var/lib/dpkg/status

Howdy, and thanks for helping to improve Mythbuntu. usplash terminating early obviously is not good, but it affects all Ubuntu distributions, not just Mythbuntu. Since it isn't something that Mythbuntu team can do anything about, I'm going to mark as invalid for that project. It will simply pick up that fix if/when it is resolved for all of Ubuntu.

From the description and comments it does not look like Upstart is doing anything wrong here, though this may be a bug with the initscripts that run fsck

I have the following, which might be related to the same thing, although it's a different problem:

I have 3 additional partitions which are mounted using fstab on boot.
When a fsck runs on one of those partitions, gdm starts while fsck is still running. The partitions are not mounted then.
Mount -a works after fsck is finished.

I'm also running two home "servers", both have a second partition for video-recordings, backups, virtual machines and they both show the same problem. When a fsck runs on the second partition, system does not wait unitl it finishes but boots up. Programs like vdr, vmware-server and backuppc terminate because "their" mount point is empty.
mount -a and a restart of the affected services brings it up again.

I've got no problem with fsck at boot with karmic. I think this is related to the increasing activation of upstart instead of sysvinit.

I think there won't be a fix because ubuntu switches to upstart, which is not affected by this bug.