On Thu, Jun 5, 2014 at 6:57 PM, Vincent Lasmarias
<vlasmarias@contigo.com> wrote:
> Thanks for the informative responses and suggestions. My responses below:
>
> * Sorry for the double post. I posted the original message using my gmail
> account and got a "is not a member of any of the restrict_post groups"
> response and when I didn't see it for a day, I ended up wondering if it was
> due to my use of a gmail account - so I tried using my company email account
> instead to post an updated version of the original post.
>
> * Our system is not virtualized.
>
> * Jeff, the output format of the load and free/cached memory did not come
> from a tool but came from my script. My script does OEuptime; free m¹, and
> then another script massages the data to only grab date, 1-minute load
> average, free, and cached.
>
> * For the 'top' outputs, I don't have the actual 'top' output, but I have
> the following:
All right; top is pretty clearly indicating a problem in the kernel.
Maybe this is numa or something else. If you can reproduce this on
the 'newer' kernel, it might be useful to try and get a 'perf' going;
it was introduced in 2.6.31. This might fail for one reason or
another though but if you could get it rigged it might give some clues
as to where in the kernel you're getting bound up.
Long term, looking into upgrading the O/S might be a smart move.
merlin
