On Fri, Mar 13, 2009 at 4:13 PM, David Vasek <[email protected]> wrote:
It is wrong becasue the computed numbers can be different from what is written
in the specification (the man pages). The computed load average can be high on
an almost idle machine and vice-versa. As is described here:
http://marc.info/?l=openbsd-misc&m=118703405121404
http://marc.info/?l=openbsd-misc&m=93551115818166
http://marc.info/?l=openbsd-misc&m=121849543013236
I guess that this is the cause for all the repeated discussions about high load
average. It can't be fixed without redesigning large portion of the kernel, if
it can be fixed at all and it would definetely be for some performance
trade-off.
One exception here. You use this one:
> http://marc.info/?l=openbsd-misc&m=121849543013236
As a way to say the load average calculation is/was wrong, but if you
ready it more, it's not related to the load average calculation as far
as I can tell with more research on the issue. But in my humble opinion
with way more work on this then I care to say to be honest as I still do
not have the answer after a year of looking on/off into it that the
issue in this particular case is related with the USB driver some how
that always try to see if there is something connected to it.
Proof of this is you disable the driver
# config -e -u -o /nbsd /bsd
OpenBSD 4.5 (GENERIC) #1898: Sat Feb 28 17:42:44 MST 2009
[email protected]:/usr/src/sys/arch/sparc64/compile/GENERIC
History is empty
Enter 'help' for information
ukc> disable 240
240 usb* disabled
ukc> quit
Saving modified kernel.
# rm /obsd ; ln /bsd /obsd && mv /nbsd /bsd
# chmod 644 /bsd
# reboot
And this is now gone for good.
Or an other way, you plug a USB device into the UBS port if memory
served me well when I tried last year and you have the same results.
The error by one I explained there is not an error in the calculation,
but really the plug and play if you want for the lack of better words
that check the USB all the time that run constantly, witch I think
shouldn't be affected like that and somehow get stuck in a loop or
something as you reboot the same boxes many times and 1 out of 10, you
will not have that issue.
So, as a practice work around for now until this can be isolated, I
simply disable the driver 240 in the kernel and all is good and have
been good for a year so far on almost 200 Sun V100, and X1 servers.
Not to confuse the issue here.
Just FYI and I hope it correct that as far as the last reference in your
posting point too. If you read more on that tread, you will see that as
well as I corrected the first statement I did on the subject.
Best,
Daniel
