OpenVZ 2.6.18-ovz028test010.1 crashing on DELL PE 2950 [message #9735] |
Fri, 19 January 2007 10:20 |
Jan Tomasek
Messages: 44 Registered: December 2006
|
Member |
|
|
Hello,
later I was reporting problems when running 2.6.18-ovz028test005.1 on my
DELL PE 2950. Very similar problem came back with 2.6.18-ovz028test010.1.
HW description:
I'm trying to run OpenVZ on my new DELL PE 2950, it has 8GB of RAM, 2x
Intel Xeon 5160 @ 3.00GHz. 1.1TB FS attached to internal Perc 5/i
adapter and 5.9TB FS from MD1000 attached to Perc 5/E. BIOS and Firmware
of both SAS controlers was updated to latest versions.
After instalation of 2.6.18-ovz028test010.1 I didn't any test, it was
working fine until Thursday morning when masive I/O started. System
stoped completely responding. Screen shows this:
http://www.tomasek.cz/stuff/IMG_0642.JPG
system was not respnding to anything.
After fresh reboot I started simulating masive I/O by
while true; do nice -20 dd if=/dev/zero of=/sklad/`date +"%s"` bs=1G
count=4; done
and in while system started be unresponsive. It does not crashed as
previous, even ping was responding, but that was all. All shells to
system was dead. Alt+SysRq+... printed memory info and cpu registers:
http://www.tomasek.cz/stuff/P1070747.JPG
http://www.tomasek.cz/stuff/P1070748.JPG
proceces were not respoinding to term/kill signals. I had to reboot system.
I was interested in 2.6.18-ovz028test010.1 because of it's ability log
iptables mesages into VE logs.
Now I'm back running 2.6.18-ovz028test007, it works without any problems.
If I can provide more info please ask me, I will try to help. I learned
how to make that font smaler so more info could fit on screen. I also
can disable that blaking screen which doing something in all crashes I
see with OpenVZ.
Best regards
--
-----------------------
Jan Tomasek aka Semik
http://www.tomasek.cz/
|
|
|
|
Re: OpenVZ 2.6.18-ovz028test010.1 crashing on DELL PE 2950 [message #9782 is a reply to message #9736] |
Mon, 22 January 2007 09:05 |
Jan Tomasek
Messages: 44 Registered: December 2006
|
Member |
|
|
Hello Pavel,
Pavel Emelianov wrote:
>> After instalation of 2.6.18-ovz028test010.1 I didn't any test, it was
>> working fine until Thursday morning when masive I/O started. System
>> stoped completely responding. Screen shows this:
>> http://www.tomasek.cz/stuff/IMG_0642.JPG
>> system was not respnding to anything.
>
> This is most likely the tail of the lockup that can be seen
> on the third screenshot.
>
> We've found a race in IO accounting in 028test010 kernel.
> You may pull the latest git tree or just apply a single
> patch to 010 kernel. This must help.
>
> The patch URL is
> http://git.openvz.org/?p=linux-2.6.18-openvz;a=commit;h=563172f5c9a1023465819505ac17bdd8dab2328f
I patched kernel on friday, it survived intensive testing and whole
weekend. Thanks! :)
--
-----------------------
Jan Tomasek aka Semik
http://www.tomasek.cz/
|
|
|