On Mon, Dec 8, 2008 at 10:47 AM, Liraz Siri <liraz@turnkeylinux.org> wrote:
>> This sounds great. If I wind up with a big machine to test it on I'll
>> tell you how it goes.
>
> Thanks, though you don't necessarily have to wait for a "big machine" if
> you just want to give the appliance a go. A little machine or even a
> virtual machine should do.
Oh, I'll definitely try it on smaller machines. but...
>> Are you familiar with this bug:
>>
>> https://bugs.launchpad.net/debian/+source/linux/+bug/245779
>>
>> It's the reason my latest db servers are running Centos 5.2, sadly.
>> By the time I'd found the suggested workaround of setting a boot
>> option of NO_HZ=y I was already migrated off ubuntu for db servers.
>
> No, I wasn't familiar with the bug. Very strange. I've been using Hardy
> with that kernel for a few months and have yet to come across it...
Well, it only seems to show up on certain high performance hardware
setups, i.e. many cores, heavy access, etc. At least for me. We run
pgsql on a number of different boxes on ubuntu 8.04 and 7.10 with no
real problems. but on that one machine with 8 opterons and 16 disks
on a fast raid controller, it shows up within a few days of heavy use,
stealing one CPU at time until the machine locks up.