[Lula] system load averages

[Steve Glasser]

Well, we got way too many boxen, not enough power or cooling, you can
figure out the rest... ;)

On 4/8/07, Peter Benjamin <pete at peterbenjamin.com> wrote:
> At 06:55 PM 4/8/2007, you wrote:
> >On 4/8/07, Steve Glasser <steve at fpig.net> wrote:
> >> Both 'top' and 'uptime' produce numbers for system load averages at
> >> one, five and fifteen minutes.  I've read a bunch of stuff about
> >> system load averages but it seems vague and contradictory.  What
> >> numbers would actually indicate that the system cpu was overloaded?
> >
> >Anything higher than the total number of cpu cores.
> >So a dual CPU system should never show a load average over 2.0.
> >If it does, it means that more jobs are running than there is CPU for.
>
> Oh, good answer Dan.  Better than mine in degree of specific minimum value.
>
> Which is a condition that should clear itself, within
> microseconds, and if it does not, then do worry, well
> at least test it again in a few seconds.
>
> Typically, I've found when it goes over the CPU count, and
> continues to grow, it's a problem that needs immediate
> intervention.  Or at least being sat on, and looked at
> and frowning a lot, knowing there are no processes to
> cancel to clear the condition, as all processes are
> critical...
>
> Always tell the boss, so when he gets a phone call complaining
> about it, he can tell the caller "We know, and are looking into it",
> thus, making the boss look good, compared to him coming to you
> and saying "What the f_ck?", and having to call his boss back
> with an estimated resolution time.
>
> And telling the boss allows him to track the monthly, then
> weekly, then daily the impact, and allows scheduling of optimizations
> or some type of upgrade, or moving of services, to other machines,
> instead of the in one day Rush to clear the condition, in crisis
> management mode.  Your boss... YMMV how he responds.  At the
> very least you can say CYA, particularly when it became weekly
> you memo'ed him a graph, in writing, of the problem, and the
> estimated time to crashing, and desired solution direction,
> and time frame of solution, and when the solution needs to
> start being implemented.
>
> Which leads to, hey, Steve, why do you want to know?
>
>
> _______________________________________________
> Lula mailing list
> Lula at lula.org
> http://lula.org/mailman/listinfo/lula_lula.org
>


-- 
Steve Glasser
Flying Pig Computer Svc.
sgla9347 at gmail.com