[nzlug] Load stuck at 1

[Nick 'Zaf' Clifford]

Steve Holdoway wrote:
> If the processes are very short-lived, then they may only rarely get caught. I have a mail server ( sendmail ) that regularly reports load averages of 250+, on a dual dual-core machine. Works fine - always responsive. So, if you get a glimpse, now and again, of a process using one of your cores, it's probably that one all of the time.
>   
Ummm... no.
By definition.
Load averages are, as they say, the average system load for the past 1,
5 and 15 minutes.
If you have a system with a 15 minute load average of 250, you have
something very, very, broken (or an extremely busy system).

Load average is not instantaneous, this is why it takes time for the avg
to come down after you kill firefox cause its chewing at 100% again.

Lets define how the load average is calculated, simplistically its the
exponential moving average of the number of processes waiting in the
run-queue, plus the number of currently running processes, averaged over
the 1,5 and 15 minute intervals.

If you have a load average of 1, it means the CPU is being used 100% of
the time (if you were to start another process, it would have to wait
before it could be run). A load average of 1.00 is a good thing if you
are a server (optimum efficient use of resources), and a sucky thing if
you're a laptop (power/battery conservation).

Of course, this explanation ignores some of the trickyness, eg a task
stuck in an uninterruptible sleep will raise the average.

As for what causes load average stuck at 1, I've seen it on some
distros, (I think SuSE), and I think, but I'm not 100% sure, its a
kernel issue.
Try this article: http://kerneltrap.org/Linux/High_Idle_Load_Average

Nick