F9, F8 periodicially hanging upon user login, dbus zombie process, etc.

[Cameron Mura]

Hi all,

Still haven't been able to solve the F9 locking problem (email below), 
part of the delay being that I don't know what exactly triggers it (so 
can't easily test it, and have to wait some time (~ 1 day) before it 
falls into this lock-able state...).

But, in case it's of any help to anyone who's seen anything like this or 
has any thoughts about it, I notice that the problem seems to arise 
sometime after the ~ 4 AM cron tasks (cron.hourly, daily, etc.).  Here's 
what's in the hourly and daily (i.e., minimally at the frequency of 
occurrence of this problem):
> cron.daily
> cron.daily/mlocate.cron
> cron.daily/0anacron
> cron.daily/readahead.cron
> cron.daily/makewhatis.cron
> cron.daily/cups
> cron.daily/certwatch
> cron.daily/texlive.cron
> cron.daily/logrotate
> cron.daily/0logwatch
> cron.daily/prelink
> cron.daily/000-delay.cron
> cron.daily/00webalizer
> cron.daily/tmpwatch
> cron.daily/rpm
> cron.deny
> cron.hourly
> cron.hourly/drupal
> cron.hourly/mcelog.cron
(Much of this is unnecessary (like drupal) for most of the workstations, 
but my approach in constructing a kickstart for these has been to create 
rather fat clients and then trim down to get the final kickstart config...)

Also, I just reproduced the problem for today:

1)  Actively logged-in to a GNOME session as root

2)  Navigate to the menu bar -> logout

3)  Screen freezes black, with the mouse pointer still active (and the 
blue elliptical "actively working" graphic circulating about the pointer)

Remotely log-in to the frozen machine via ssh, see the dbus-related zombie:
> [root at struc12 etc]# ps waux |grep Z
> USER       PID %CPU %MEM    VSZ   RSS TTY      STAT START   TIME COMMAND
> gdm      28654  0.0  0.0      0     0 ?        Zs   10:30   0:00 
> [dbus-launch] <defunct>

It seems to be a gdm thing, as I can switch to a terminal -- for 
example, ctrl-alt-F6 works.

...only remedy is to re-boot.

Any thoughts, advice, or tips on how to go about troubleshooting this 
would be greatly appreciated.

With best regards,
Cameron



=== Cameron Mura wrote (on 08/10/2008 03:05 PM): ===
> Hello,
>
> I'm experiencing a problem on both F8 and F9 (mostly playing on F9 
> now) that I can't seem to get to the bottom of: In several different 
> instances of vanilla installations of both F8 and F9 on Dell 
> workstations (Precision 490, T3400, etc.), I find that the machine 
> works fine for about 1 day, then -- the next day -- hangs at a black 
> screen when a user tries to login. This is with both i386 and x86_64 
> installs.
>
> Gnome is the default disp. manager, and if I remotely access the 
> machine (ssh as root) while it's in this frozen state, I see that 
> dbus-launch (user 'gdm') has fallen into a zombie state.  Re-booting 
> solves the problem, but then it arises again the next day.  I should 
> mention that the regular (non-root) user accounts are set-up via 
> NIS/NFS. (I ran across some things online regarding problems with 
> pulseaudio and nfs users, but this doesn't seem to be it.).   Other 
> suspicious behavior that seems to (maybe) be correlated is that 
> launching into a bash shell (for, say, a tcsh-based user) results in a 
> stream of infinite /dev/null errors (don't have the exact output 
> string in front of me, but can provide that later if helpful)... This 
> can be interrupted (ctrl-c) to yield the shell.  Further inspection 
> shows that the perms on /dev/null are only rw for root (not 0666 or 
> whatever they ought to be...?).  I don't know if these two things (1: 
> login screen locking/zombie process 2: screwy /dev/null perms) are 
> related.
>
> I'm wondering if anyone's seen anything like this, or might have some 
> idea as to what's causing it? If so, please let me know, or feel free 
> to redirect me to any info about this online (I've spent many hours 
> google'ing for it, tweaking config files, applying different sets of 
> updates, etc...).
>
> Thanks for any help,
> Cameron
>
>
> PS. This is also at 
> http://forums.fedoraforum.org/showthread.php?p=1061369 . Any help / 
> tips / advice would be most gratefully appreciated.
>

-- 

Cameron Mura

-------------- next part --------------
An HTML attachment was scrubbed...
URL: <http://listman.redhat.com/archives/fedora-list/attachments/20080811/872d7d7e/attachment-0001.htm>