Re: Interactivity issue in 2.6.25-rc3

[Carlos R. Mafra]

On Fri 29.Feb'08 at 17:04:08 +0100, Ingo Molnar wrote:
> 
> (on-list)
> 
> * Carlos R. Mafra <crmafra2@xxxxxxxxx> wrote:
> 
> > Is it an scheduler anomaly if 'se.wait_max' is bigger than 40 msecs 
> > for _any_ of the processes which appear in the debug script log? In 
> > other words, is the scheduler mathematically build to not allow 
> > latencies higher than 40 msecs?
> 
> it is definitely an anomaly if it's bigger than 40 msecs if you clear 
> all stats via cfs-debug-info-clear.sh and the large latencies appear 
> after that. You can force it to go above 40 msecs if you run more than 
> say 40 CPU hogs in parallel, so it's not "mathematical", but you should 
> never see large latencies under normal workloads - and that includes 
> almost everything but "insanely high" workloads.

Thank you for the explanation! 

> and obviously, even if you only 'feel' long delays that's too an anomaly 
> by definition, no matter what the scripts say about it. It might even be 
> a scheduler anomaly as well: for example if the scheduler clock has an 
> anomaly - on which the delay statistics are based too.

But if the scripts say all 'se.wait_max' are < 40 msecs than it is
not CFS' fault, right? Even if it takes 3 seconds for a typed letter
to appear in the terminal?

> generally, latencytop gives a pretty good idea about where app delays 
> come from. (As a second-level mechanism, in sched-devel.git you can try 
> the latency tracer.)

Yeah, I must try latencytop to check for more things before sending
an email reporting possible problems. 

Thanks again!
--
To unsubscribe from this list: send the line "unsubscribe linux-kernel" in
the body of a message to majordomo@xxxxxxxxxxxxxxx
More majordomo info at  http://vger.kernel.org/majordomo-info.html
Please read the FAQ at  http://www.tux.org/lkml/