https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c2 Joachim Banzhaf changed: What |Removed |Added ---------------------------------------------------------------------------- Priority|P5 - None |P1 - Urgent Version|Factory |Final --- Comment #2 from Joachim Banzhaf 2011-04-11 17:40:34 UTC --- Just wanted to add: I did an update from 11.3 while the OP did an install: seems to have no influence I use LVM for the relevant partitions while the OP does not -> no influence The RUN_PARALLEL="no" workaround worked for me too. -> I guess some init script dependency breakage. But I don't know how to debug or fix that. Mainly because I don't know what the correct dependencies would be. Updated version to final (not factory) bacause thats what i use and it sounds like the OP uses this too. Updated priority according to wiki to urgent (data loss can/will happen). -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c4 --- Comment #4 from Peter Conrad 2011-04-14 09:56:55 UTC --- I think I found a race condition using strace on blogd + startpar. This is what happens: * blogd creates /dev/blog * /etc/init.d/halt runs /etc/init.d/rc5.d/K* * (for some reason) blogd receives SIGSYS * blogd continues reading from stdin + /dev/blog for max. 3 seconds * meanwhile, /etc/init.d/halt continues to execute startpar, which opens /dev/blog for writing * blogd closes /dev/blog * startpar tries to write to the previously opened /dev/blog, receives SIGPIPE + exits Making /etc/init.d/halt sleep for 1-2 seconds before executing startpar fixes the race. Catching + ignoring SIGPIPE should work, too. Dunno what's best. Btw, this was tested with the latest sysvinit-tools-2.88-37.45.1.x86_64 . -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c6 Markus Abt changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |abt@comet.de --- Comment #6 from Markus Abt 2011-04-15 20:42:00 UTC --- I'm experiencing the same problem. However, I'm using ext2 file-systems. Also, sometimes (ca. 50%?), the umount is executed during shutdown/reboot, and sometimes not. The RUN_PARALLEL=no work-around works on my system, too. The "sleep 2" trick (comment #4) may work, too. But I have only rebooted 3 times so far... I have put debugging statements into the /etc/init.d/halt and all used boot.* scripts, which write their output to some file in the root fs (since /var/log/boot.msg gets unavailable in the case unmount of /var succeeds). In "parallel" mode, I can see that /boot/init.d/halt continues after the startpar statement, but some of the boot.* script started by startpar finish _later_ than startpar. Also, the number of boot.* scripts called varies on every shutdown/reboot. Looks like startpar dies, as Peter has pointed out. I'm using openSUSE 11.4 x86_64, with current updates: sysvinit-2.88-37.45.1.x86_64 insserv-1.14.0-20.24.1.x86_64 Maybe the bug should be renamed since it does not relate to reiserfs. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c7 --- Comment #7 from Markus Abt 2011-04-17 18:52:24 UTC --- (In reply to comment #4)

* (for some reason) blogd receives SIGSYS

The signal is intentionally sent by the /etc/init.d/rc script: <snip> # # Do never call startpar for single, halt or reboot script # case "$RUNLEVEL" in S|0|1|6) DO_CONFIRM=no RUN_PARALLEL=no killproc -SYS /sbin/blogd esac </snip> After that, rc starts the scripts for entering the new run level, e.g. /etc/init.d/rc0.d/halt in case of a shutdown. (Note that, at the top of the halt script, RUN_PARALLEL is re-read from /etc/sysconfig/boot and therefore possibly reset to RUN_PARALLEL=yes.) The SIGSYS informs blogd that it should stop writing to disk. However, I have noticed that blogd continues to write to /var/log/boot.(o)msg for some (short) time. Typically, the last message in my /var/log/boot.omsg is "Running /etc/init.d/halt.local done". In the case that some shutdown scripts like unmounting the disks are skipped, I can see some further messages in /var/log/boot.omsg which come from startpar. It looks like this is happening in that particular case: * rc sends SIGSYS to blogd * rc calls halt * halt calls halt.local and does some other stuff * halt calls startpar * startpar calls some of the rc0.d/K* scripts * blogd receives/handles SIGSYS and stops writing to disk * my assumption is: at this point startpar exits with 141, leaving some of the rc0.d/K* scripts uncalled. The best solution for me is to set RUN_PARALLEL=no *in the halt script*. I cannot notice any negative effect on the speed for shutting down. Booting the system still uses RUN_PARALLEL=yes. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c9 Volker Kuhlmann changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |volker3204@paradise.net.nz --- Comment #9 from Volker Kuhlmann 2011-06-12 07:10:21 UTC --- I see the same problem. Most of the time, / and /home are not cleanly unmounted. I use reiserfs there too, but doubt that the type of file-system is relevant to this problem. During boot, kilometers of journal replays scroll past, esp when not having rebooted for a few days. This is not really good; it's about the same as just pulling the power plug for each reboot. My system configuration is /dev/md0 / reiserfs ... /dev/md1 /home reiserfs ... Both are kernel raid1. oS 11.4 final, all updates installed. Worked fine with oS 11.1. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c zj jia changed: What |Removed |Added ---------------------------------------------------------------------------- CC| |zjjia@novell.com AssignedTo|bnc-team-screening@forge.pr |ro@suse.com |ovo.novell.com | -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c12 Dr. Werner Fink changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |NEW CC| |werner@novell.com InfoProvider|werner@novell.com | --- Comment #12 from Dr. Werner Fink 2011-08-01 13:56:03 UTC --- (In reply to comment #11) No I was not as I had nerver reports about SIGPIPE and startpar. Beside this on SIGSYS the blogd does not write anymore to disk as then the logging file struct will be closed. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c14 Ruediger Oertel changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEW |ASSIGNED AssignedTo|ro@suse.com |werner@novell.com --- Comment #14 from Ruediger Oertel 2011-08-01 15:09:48 UTC --- so this sounds like it is not aaa_base but sysvinit, reassigning to maintainer -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c16 --- Comment #16 from Dr. Werner Fink 2011-08-01 15:55:29 UTC --- I'd like to suggest http://download.opensuse.org/repositories/Base:/System/openSUSE_11.4/x86_64/ this should do the job -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c17 Markus Abt changed: What |Removed |Added ---------------------------------------------------------------------------- Status|NEEDINFO |ASSIGNED InfoProvider|abt@comet.de | --- Comment #17 from Markus Abt 2011-08-02 13:29:53 UTC --- (In reply to comment #16)

I'd like to suggest

http://download.opensuse.org/repositories/Base:/System/openSUSE_11.4/x86_64/

this should do the job

After 20 successful halts/reboots, I can confirm that the problem is solved on my opensuse 11.4 x86_64. startpar exits with 0 now, after calling all boot.*(shutdown) scripts. I can also confirm, that blogd now stops writing to disk after SIGSYS. (FYI, blogd seems not to close /var/log/boot.msg after that, but in boot.localfs, mkill kills blogd, therefore umount /var is also successful.) ((in one situation, umount did not happen: when going from a multi-user runlevel to runlevel 1, then runlevel 0 or 6. See bug #709825.)) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c19 Dr. Werner Fink changed: What |Removed |Added ---------------------------------------------------------------------------- Status|ASSIGNED |RESOLVED Resolution| |FIXED --- Comment #19 from Dr. Werner Fink 2011-09-02 15:53:10 UTC --- opps this one is fixed -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c21 --- Comment #21 from Dr. Werner Fink 2011-09-05 07:30:05 UTC --- IMHO this updates should be already part of your system, shouldn't it? Compare with your comment #17 -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c23 --- Comment #23 from Dr. Werner Fink 2011-09-05 10:00:30 UTC --- Hmmm ... strange as I've never seen this problem, OK I've enabled automatic update within YaST and discovered onyl Firefox as a problem due the new upstream version numbering scheme, but this is not caused by SUSE ;) -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c25 --- Comment #25 from Dr. Werner Fink 2011-09-05 10:38:19 UTC --- Yes I`ve added the repository and yes the I'm able to work with :] -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.

https://bugzilla.novell.com/show_bug.cgi?id=679671 https://bugzilla.novell.com/show_bug.cgi?id=679671#c27 --- Comment #27 from Markus Abt 2012-03-16 18:31:23 UTC --- I cannot test the latest update, since I'm not using openSUSE 11.4 any longer. However, you may try this work-around: Return to RUN_PARALLEL=yes for booting, and modify /etc/init.d/halt: 1. Find the line containing "$COLD_BOOT". 2. Above this line, insert: RUN_PARALLEL="no" Additionally, you may check, if the test on "COLD_BOOT" is correct (see #709825): if test "$COLD_BOOT" != "yes" ; then Note, that any future updates may remove this modifications and bring you back the official halt script. -- Configure bugmail: https://bugzilla.novell.com/userprefs.cgi?tab=email ------- You are receiving this mail because: ------- You are on the CC list for the bug.