How to obtain which interrupts cause system to hang?

[Коньков Евгений]

Здравствуйте, b..

Вы писали 9 октября 2010 г., 21:45:07:

>> How to obtain what nasty happen, which process take 36-50% of CPU
>> resource?

bf> It partly depends upon the version of the OS that you are running,
bf> your hardware, and your configuration. 2008 interrupts/sec is high,
bf> but not improbably so, for kern.hz>=1000, and not beyond the
bf> capabilities of today's average computer, although it may not be
bf> optimal for your workloads.  So it may not actually be a problem with
bf> too many interrupts, but rather with some erroneous system accounting.
bf>  This can happen, for example, if your computer is using some of the
bf> deeper power-saving C-states, but using them poorly, as for example
bf> in:

bf> http://lists.freebsd.org/pipermail/freebsd-current/2010-August/019383.html

bf> So you may want to examine, and perhaps adjust, your choice of kernel
bf> timers, kern.hz, and power-saving settings.  Device polling can also
bf> be a factor, if you are using it.

system is one processor

vpn_shadow# uname -a
FreeBSD vpn_shadow.in 7.3-RELEASE-p2 FreeBSD 7.3-RELEASE-p2 #3: Mon Jul 26 01:56:03 EEST 2010     kes at in:/usr/obj/usr/src/sys/vpn  amd64

kern.hz is 2000, Interrupts 1984 total is ok
polling is enabled. Server runs about 6month and...
and now it has interrupts: 13.5% system, 47.7% interrupt

I can not imaginge what happen, in /var/log/* is silent... all is ok as usual.

Network while that is happening is unstable. ping on 100Mbit link is 60ms.

-- 
С уважением,
 Коньков                          mailto:kes-kes at yandex.ru