upstart

Bug #1073433
Comment #94

Comment 94 for bug 1073433

Revision history for this message

Alexander (lxandr) wrote on 2013-04-11:

#94

mega-log-1365656098 Edit (27.5 KiB, text/plain)

Max, as I've understood, "no splash" doesn't disable plymouth. Some people suppose that the freezed plymouth is the cause:
https://bugs.launchpad.net/ubuntu/+source/upstart/+bug/1019347
I've decided to find it out and totally purged plymouth from system. But that didn't help.

So, what do we have now?
I've modified /etc/init.d/umountroot script (part at 86 line):
=========================
for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
do

                mount $MOUNT_FORCE_OPT -n -o remount,ro -t dummytype dummydev / 2>/dev/null \
                || mount $MOUNT_FORCE_OPT -n -o remount,ro dummydev / 2>/dev/null \
                || mount $MOUNT_FORCE_OPT -n -o remount,ro /
                ES=$?
# [ "$VERBOSE" = no ] || log_action_end_msg $ES
                log_action_end_msg $ES
                if [ $ES -gt 0 ]
                then
                    LOG=/mega-log-`date +%s`
                    echo "`date`: Attempt ${i}: remount rootfs failed! Trying again..." #| tee -a $LOG
                    echo "`date`: Attempt ${i}: remount rootfs failed! Trying again..." >> $LOG
# sleep 1
                    { cat /proc/mounts ; echo ; ps aux ; echo ; /usr/bin/lsof ; echo ; } >> $LOG
                    /bin/fuser -v -m -u / 1>/dev/null 2>> $LOG
                    sleep 1
                else
                    break
                fi
        done
=========================

Today's morning I've got a 21 message "/ is busy" at shutdown.
Early, I thought that the cause is that some processes are accessing the disk during shutdown (writing). Some processes which are not terminated on time, so " / is busy " message appears.
Now I'm looking into logs and... Well, looks like it isn't the real cause.
I don't see any files opened for writing (except pipe, @/com/ubuntu/upstart, /dev/null and /dev/console).
So, more and more I think, that this bug is in 'init/upstart'... or in kernel!
It's not so easy to reproduce it. I've tried a lot of times to reboot machine under heavy cpu and I/O load and bug doesn't apper (a lot of times in about 10-15 minutes).
But bug appears more frequently when uptime is high enough (at least 1/2 of a day).
Steps to reproduce: push power button from running X session.

One more interesting thing: my root partition is on usb hard drive.

If I'll have some more free time near future, I'll try to test it with a new kernel ( > 3.5).

So, what do we have now?
I've modified /etc/init.d/umountroot script (part at 86 line):
=========================
        for i in 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21
        do

mount    $MOUNT_FORCE_OPT -n -o remount,ro -t dummytype dummydev / 2>/dev/null \
                || mount $MOUNT_FORCE_OPT -n -o remount,ro              dummydev / 2>/dev/null \
                || mount $MOUNT_FORCE_OPT -n -o remount,ro                       /
                ES=$?
#       [ "$VERBOSE" = no ] || log_action_end_msg $ES
                log_action_end_msg $ES
                if [ $ES -gt 0 ]
                then
                    LOG=/mega-log-`date +%s`
                    echo "`date`: Attempt ${i}: remount rootfs failed! Trying again..." #| tee -a $LOG
                    echo "`date`: Attempt ${i}: remount rootfs failed! Trying again..." >> $LOG
#                    sleep 1
                    { cat /proc/mounts ;  echo ; ps aux ; echo ; /usr/bin/lsof ; echo ; } >> $LOG
                    /bin/fuser -v -m -u / 1>/dev/null 2>> $LOG
                    sleep 1
                else
                    break
                fi
        done
=========================

Today's morning I've got a 21 message "/ is busy" at shutdown.
Early, I thought that the cause is that some processes are accessing the disk during shutdown (writing). Some processes which are not terminated on time, so " / is busy " message appears.
Now I'm looking into logs and...  Well, looks like it isn't the real cause.
I don't see any files opened for writing (except pipe, @/com/ubuntu/upstart, /dev/null and /dev/console).
So, more and more I think, that this bug is in 'init/upstart'... or in kernel!
It's not so easy to reproduce it. I've tried a lot of times to reboot machine under heavy cpu and I/O load and bug doesn't apper (a lot of times in about 10-15 minutes).
But bug appears more frequently when uptime is high enough (at least 1/2 of a day).
Steps to reproduce: push power button from running X session.

One more interesting thing: my root partition is on usb hard drive.

If I'll have some more free time near future, I'll try to test it with a new kernel ( > 3.5).