28 Mar
2011
28 Mar
'11
11:10 a.m.
Hi Adam,
On Mon, Mar 28, 2011 at 10:05:43AM +0100, Adam Sweet wrote:
I don't expect it to fix what you're seeing, but on Lenny I would
also recommend clocksource=jiffies. (not for pvops kernels like in
squeeze though)
$ cat /sys/devices/system/clocksource/*/current_clocksource
jiffies
Again I wouldn't expect this to fix anything, but it's worth a try
because it's so easy. The down time would only be on the order of
10-15 minutes. Please send a support ticket if you'd like to try
this, giving some ideas of what sorts of times you'd be OK with it
happening.
Cheers,
Andy
--
"I'm /extremely/ miffed about today's events and in my quest to try to make
you understand the level of my unhappiness, I'm likely to use an awful lot
of what we would call /violent sexual imagery/ and I just wanted to check
that neither of you would be terribly offended by that." -- Malcolm Tucker
On 27/03/11 11:32, Andy Smith wrote:
The kernel hasn't paniced as such. What's happened is a task has been
stuck inside the kernel for 120 seconds or whatever and this
automatically produces the backtrace you saw.
If it's always apache2 then it could actually be apache using all
the memory and going into swap, then being too slow to do anything.
It might be something else though. Or it might not be memory-related
at all.
Sorry, this is a new one to me. Is it always
apache2 task that gets
blocked?
Yes, always. On previous occurrences I did see that the OOM handler had
killed apache2 and mysqld, but I didn't expect the kernel to panic and
the system to become inoperable in this way. What are your
kernel boot arguments?
kernel /boot/vmlinuz-2.6.26-2-xen-686 root=/dev/xvda1 console=hvc0 ro We could try
moving your VPS to another host. Probably worth trying
that before migrating/reinstalling.
I'd be happy for you to try that, I'd prefer to arrange a time out of
office hours so I can let people know that their mail won't work during
the move.