Date: Tue, 6 Sep 2011 20:32:55 +0300 From: Alex Kozlov <spam@rm-rf.kiev.ua> To: Poul-Henning Kamp <phk@phk.freebsd.dk>, freebsd-current@freebsd.org, spam@rm-rf.kiev.ua Subject: Re: Unusually high LA without any load at FreeBSD9-BETA2 Message-ID: <20110906173255.GA96949@ravenloft.kiev.ua>
next in thread | raw e-mail | index | archive | help
On Tue, Sep 06, 2011 at 05:22:02PM +0000, Poul-Henning Kamp wrote:
> In message <4E66547D.2030907@cran.org.uk>, Bruce Cran writes:
>>On 06/09/2011 18:05, Garrett Cooper wrote:
>>> What is "LA"?
>>Load Average?
> We should kille the load avarage as a measure for system activity,
> it only has any relevance if you run heavy CPU bound processes.
It may be true, but in current kernel from beginning of june I would
see la about 0,01 in this situation.
Note that cpu idling:
dev.cpu.0.freq: 300
dev.cpu.0.freq_levels: 2200/35000 1925/30625 1650/26250 1600/23000 1400/20125 1200/15000 1050/13125 900/11250 750/9375 600/7500 450/5625 300/3750 150/1875
dev.cpu.0.cx_supported: C1/1 C2/1 C3/17
[...]
11 root 2 155 ki31 0K 32K CPU1 1 29.4H 200.00% idle
I think process accounting get broken or something like this.
> If the majority of your threads yield their quantum, load average
> contains absolutely no information of any relevance to system
> capacity.
>
> Try this:
>
> main()
> for (i = 0; i < 10000; i++)
> start thread {
> calculate time until top of next second
> sleep (until then)
> }
>
> You'll see a monster load-avg on idle cpus.
--
Adios
Want to link to this message? Use this URL: <https://mail-archive.FreeBSD.org/cgi/mid.cgi?20110906173255.GA96949>