[regression][lucid] system frequently reboots after update in august 2012, path via rtkit and AptDaemon

[jan]

My two lucid (LTS- long term support) systems now display recent reboots.
I suspect the kernel upgrade that was released end august 2012, but I an not sure.
During the time of the reboot, the system was doing calculations in and atd/cron job.
The signature of the reboots is as follows:
1) rtkit-daemon starts logging that some processes become RT.
2) after about 1 minute AptDaemon: INFO: Initializing daemon
3) after about 5 minutes AptDaemon: INFO: Quiting due to inactivity
4) then, after about 30 - 60 minutes, I see the system reboot.

This is a regression: I did not see this before.
I have worked around by booting with acpi=off

 uname -a
2.6.32-42-generic-pae #96-Ubuntu SMP Wed Aug 15 19:12:17 UTC 2012 i686 GNU/Linux

logging from daemon.log:

Sep 2 18:59:11 his10 rtkit-daemon[1876]: Sucessfully made thread 8119 of process 8119 (n/a) owned by '1002' high priority at nice level -11.
Sep 2 18:59:11 his10 rtkit-daemon[1876]: Supervising 12 threads of 4 processes of 4 users.
Sep 2 18:59:15 his10 rtkit-daemon[1876]: Sucessfully made thread 8134 of process 8119 (n/a) owned by '1002' RT at priority 5.
Sep 2 18:59:15 his10 rtkit-daemon[1876]: Supervising 13 threads of 4 processes of 4 users.
Sep 2 18:59:15 his10 rtkit-daemon[1876]: Sucessfully made thread 8135 of process 8119 (n/a) owned by '1002' RT at priority 5.
Sep 2 18:59:15 his10 rtkit-daemon[1876]: Supervising 14 threads of 4 processes of 4 users.
Sep 2 18:59:16 his10 rtkit-daemon[1876]: Sucessfully made thread 8137 of process 8119 (n/a) owned by '1002' RT at priority 5.
Sep 2 18:59:16 his10 rtkit-daemon[1876]: Supervising 15 threads of 4 processes of 4 users.
Sep 2 18:59:16 his10 rtkit-daemon[1876]: Sucessfully made thread 8138 of process 8138 (n/a) owned by '1002' high priority at nice level -11.
Sep 2 18:59:16 his10 rtkit-daemon[1876]: Supervising 16 threads of 5 processes of 4 users.
Sep 2 19:00:14 his10 AptDaemon: INFO: Initializing daemon
Sep 2 19:05:15 his10 AptDaemon: INFO: Quiting due to inactivity
Sep 2 19:05:15 his10 AptDaemon: INFO: Shutdown was requested

[jan (jan-ubuntu-h-i-s)]

Both systems are synchronized to a time server.
Occurrence started within a few days from installation of:
https://lists.ubuntu.com/archives/lucid-changes/2012-August/013045.html
All other updates were installed at that moment as well, so it might be another update as well.

[Sebastian Heinlein (glatzor)]

Aptdaemon shuts down itself and not the system. Perhaps the log message of aptdaemonn should "quitting"

[Sebastian Heinlein (glatzor)]

Ah, I already renamed the log message.

[jan (jan-ubuntu-h-i-s)]

If the problem was only in the log, I wouldn't have filed a bug.

As a result, the Dell system actually reboots, and the calculations (running in batch, needing 3 days) on the system are thus invalid.
Manual intervention (re-starting the calculation) is needed almost daily, rendering the computer useless.
The workaround is now to boot with kernel option noacpi or acpi=ht.

This is only at my Dell system. Another computer keeps running fine, even with the new rtkit messages in the logging.

[Sebastian Heinlein (glatzor)]

I don't see why the reboot should be tiggered by aptdaemon. You seem to be confused by the aptdaemon log message which includes the words "shutting down". But this only refers to aptdaemon itself and not the system. That is why renamed the log messages.

[Sebastian Heinlein (glatzor)]

Ah sorry. Now I am confused: I mixed this bug report with another one. You are already getting the new "qutting" log message. Why should aptdaemon cause the reboot?

[jan (jan-ubuntu-h-i-s)]

I am not familiar with the design of the aptdaemon. I just see the message each time before the reboot, and thus I think that the aptdaemon plays a role in the process.
The way I understand is that apt may request a reboot (eg after kernel upgrade) and that a sequence triggered by the rtkit message gets the aptdaemon to do something.
In my understanding, I see the rtkit message (which was not present a few months ago) as the initial trigger.

[Sebastian Heinlein (glatzor)]

Aptdaemon does not restart the system. I k ow this since I am the main developer :)

Have you already checked your system memory using memcheck at boot time?

jan <email address hidden> schrieb:

>I am not familiar with the design of the aptdaemon. I just see the
>message each time before the reboot, and thus I think that the
>aptdaemon plays a role in the process.
>The way I understand is that apt may request a reboot (eg after kernel
>upgrade) and that a sequence triggered by the rtkit message gets the
>aptdaemon to do something.
>In my understanding, I see the rtkit message (which was not present a
>few months ago) as the initial trigger.
>
>--
>You received this bug notification because you are a member of
>Aptdaemon
>Developers, which is subscribed to aptdaemon in Ubuntu.
>https://bugs.launchpad.net/bugs/1052323
>
>Title:
> [regression][lucid] system frequently reboots after update in august
> 2012, path via rtkit and AptDaemon
>
>To manage notifications about this bug go to:
>https://bugs.launchpad.net/linux-kernel-bugs/+bug/1052323/+subscriptions

--
Diese Nachricht wurde von meinem Android-Mobiltelefon mit K-9 Mail gesendet.

[jan (jan-ubuntu-h-i-s)]

A) fortunately, this dismisses aptdaemon as a part of the sequence.
I'll remove aptdaemon as package involved in this bug.
B) Not yet. I will do so later this week.